Abstract.
Early diagnosis and treatment of childhood fever, an important sign of potentially serious infections such as malaria, is essential for controlling disease progression, and ultimately, preventing deaths. This study examined individual- and community-level factors associated with treatment-seeking behaviors and promptness in these behaviors among caregivers of febrile under-five children in Malawi. The 2015–2016 Malawi Demographic Health Survey was used to analyze a nationally representative sample of 4,133 under-five children who had fever within 2 weeks before the survey. A multilevel logistic regression model was used to examine the association between individual- and community-level factors and treatment-seeking behaviors. Approximately 67.3% of the caregivers reported seeking treatment for their febrile child, whereas only 46.3% reported promptly seeking treatment. Children from communities with moderate and high percentages of educated caregivers were more likely to be taken for treatment (adjusted odds ratio [aOR] = 1.26, 95% CI = 1.01–1.58 and aOR = 1.31, 95% CI = 1.02–1.70, respectively) than those from communities with a low percentage of educated caregivers. Children from communities with moderate and high percentages of caregivers complaining about the distance to a health facility were less likely to be taken for treatment (aOR = 0.74, 95% CI = 0.58–0.96 and aOR = 0.67, 95% CI = 0.51–0.88, respectively). At the individual level, having a cough in the last 2 weeks, region, religion, and having better health behaviors in other health dimensions were associated with fever treatment-seeking behaviors among Malawian caregivers. Programs aimed at improving treatment-seeking behaviors should consider these factors and the regional variations observed in this study.
INTRODUCTION
Childhood fever is one of the typical signs of illness in young children.1 Fever is also said to be a common clinical sign of Plasmodium falciparum infection.2–4 Evidence has shown that a large proportion of fever in malaria-endemic countries may be attributed to malaria.5 For instance, a Tanzanian study reported that 71.4% of fever cases were attributable to malaria.5 Therefore, in many malaria-endemic countries, malaria is suspected primarily on the basis of fever.5–8 The WHO recommends early malaria diagnosis (within 24 hours of the onset of symptoms) to prevent progression from uncomplicated to severe malaria.9 The use of childhood fever to diagnose malaria in under-five children has aided the early diagnosis and treatment of uncomplicated malaria, thereby reducing the effects of severe malaria.10–12 In addition, childhood fever may underline other childhood illnesses such as viral or bacterial infections.13 Therefore, to timely diagnose, offer prompt management, and provide appropriate care, caregivers with febrile children should seek care at health facilities.5,14
Despite the importance of fever as an early sign of malaria and other childhood illnesses, the WHO reported that only 34% of caregivers with febrile children in sub-Saharan Africa (SSA) had sought treatment from a health professional in 2016.9 This low rate of care-seeking behavior is a concern because most SSA countries are malaria endemic. Previous studies have revealed that most malaria-related deaths occur at home, with no medical help sought in such cases.10,15 Moreover, when the treatment is sought, it is mostly delayed.10 When dealing with fever, mothers often believe that it may just be an ordinary fever, not thinking of potential malaria or other serious illnesses.16 A lot of women resort to home-based treatments for fever, which may subsequently result in severe illnesses.16 In Malawi, 49.9% and 59.0% of febrile children younger than 5 years were brought for treatment in 2012 and 2014, respectively.17–21 Although treatment-seeking behaviors among Malawian caregivers have steadily increased, the rates are relatively lower than those reported in other SSA countries such as Zambia (77.0%) and Ethiopia (87.8%).10,22 Enhancing the fever treatment-seeking behavior in Malawi is, therefore, imperative. In addition, in Malawi, rates of prompt treatment-seeking behavior have been relatively lower than those of general treatment-seeking behavior, without considering the promptness.23,24 Understanding the factors that influence treatment-seeking behaviors, especially the promptness, of caregivers in Malawi is important. Discerning these factors would help public health programs seeking to improve early diagnosis of childhood illnesses, such as malaria, to formulate targeted interventions.
Several studies have demonstrated that socioeconomic factors such as region, residence, caregiver’s age, listening to radio, and perceived barrier to seeking treatment were associated with treatment-seeking behaviors.10,25 For instance, caregivers’ exposure to the media was associated with increased likelihood of treatment-seeking behavior for their febrile children in Tanzania.26 A west Ethiopian study reported higher odds of treatment-seeking behaviors among younger caregivers (< 30 years) than among older caregivers (≥ 30 years).10 However, no association between caregiver’s age and treatment-seeking behavior was observed in a Liberian study.25 The inconsistencies in these two studies could be explained by the differences in methodological approaches. The Ethiopian study assessed age as a categorical variable, whereas the Liberian study assessed age as a continuous variable. In addition, the Liberian study considered community effects, whereas the Ethiopian study did not account for community effects.
Community socioeconomic characteristics may help mold one’s personality, attitudes, lifestyle, and health behaviors. Thus, understanding community effects is essential for designing effective programs in different communities that possess different characteristics. Few studies have examined community effects on treatment-seeking behaviors for childhood fever.26 In addition, studies examining treatment-seeking behavior have been limited in assessing the promptness at which the treatment was sought.
Therefore, using nationally representative samples, this study examined the individual- and community-level factors that influence treatment-seeking behavior of caregivers and compared the individual- and community-level factors that influence general and prompt treatment-seeking behaviors for childhood fever in Malawi.
METHODS
Study design and data source.
This was a cross-sectional study using secondary analysis of data conducted in Malawi. The 2015–2016 Malawi Demographic and Health Survey (MDHS) was used. Detailed methodology and sampling design used by the MDHS have been described elsewhere.23 Moreover, sampling involved two stages. The first stage involved the selection of 850 clusters from the 2008 Malawi Housing Population Census (the sampling frame) and household listing. The second stage involved the selection of households from the household list by using equal probability sampling criteria. A total of 30 and 33 households were chosen per urban and rural clusters, respectively. A questionnaire was used to collect information from women of reproductive age (15–49 years). The questionnaire collected information on sociodemographics and childhood illnesses. Among 25,146 eligible women, 24,562 were successfully interviewed, indicating a 98% response rate. In this study, the children’s record file (KR dataset) was used for analysis. The KR dataset contains information on children younger than 5 years whose mothers were interviewed.27 Therefore, a caregiver was defined as a mother aged between 15 and 49 years with a child/children aged 5 years or younger, who had responded to the questionnaire during the survey. The study was restricted to caregivers of febrile under-five children (n = 4,133). The frequencies among this sample revealed that each caregiver included in the analysis had only one child. There was, thus, no repetition or multiple children from the same caregiver. The unit of analysis for this study was children younger than 5 years and their caregivers.
Outcome variables.
Two outcome variables were examined in this study. First, treatment seeking was defined as whether children with fever within 2 weeks before the survey were taken for any treatment to a health facility, a pharmacy, or a formal service provider (a formal service provider included health surveillance assistants but excluded traditional healers), and assessed as a binary variable (yes/no). The second outcome, prompt treatment-seeking behavior, was defined as whether children with fever within 2 weeks before the survey were taken for any treatment to a health facility, a pharmacy, or a formal service provider within the same or the next day after the onset of symptoms.23,28–30 This was also assessed as a binary (yes/no) outcome.
Independent variables.
In the present analysis, individual- and community-level factors were selected and considered based on their importance from the literature.26,28,31–33 Key sociodemographics were further grouped to be examined at the community level.
Individual-level factors.
Individual factors included the child’s and caregiver’s individual factors, namely, gender of the child (male or female) and age of the child in months (0–5, 6–11, 12–23, or ≥ 24), whether the child had cough in the last 2 weeks (yes or no),28 whether the child had diarrhea in the past 2 weeks (yes or no),28 the caregiver’s age (15–24, 25–34, or ≥ 35 years), education (no formal education, or primary or secondary/postsecondary education), occupation status (unemployed or employed), parity (one, two, or three or more), religion (Catholics, Protestants, or Muslims and others), marital status (unmarried or married), and number of under-five children (one, two, or three or more). Three media outlets (newspapers, radio, and television) were used to measure caregivers’ media exposure. Respondents who indicated either reading newspaper, listening to the radio, or watching television at least once a week were regarded as exposed to the media.32 The variable was binary (yes or no). Other variables included region (northern, central, or southern), residence (urban or rural), and caregiver’s perceived distance to a health facility (no problem or problem). The study further assessed caregivers’ health behaviors. The number of antenatal care (ANC) visits during the last pregnancy was assessed as good (at least four) and poor (fewer than four),34 the use of postnatal care services 2 months after delivery during the last pregnancy was assessed (yes/no), and the place of delivery was assessed as good (health facility) and poor (non-health facility).23 Caregivers with all, two, and one/none of the behaviors were categorized as having good, medium, and poor health behaviors, respectively. Wealth was calculated using the principal component analysis by scoring household items and categorizing the calculated scores into quintiles from the poorest to the richest.23 The present analysis categorized the upper 40% as rich, middle 20% as the middle, and lower 40% as poor.35
Community-level factors.
In this analysis, a community was defined as a group of households sharing a common primary sampling unit/cluster within the dataset. Variables at the community level were constructed by aggregating them from individual-level factors. The community variables included community employment (proportion of caregivers in the community who were employed), community education (proportion of caregivers in the community with primary/post-primary education), community wealth (proportion of caregivers in the community who were rich), and community distance to a health facility (proportion of caregivers in the community who perceived the distance to a health facility as a problem). The continuous community-level variables were further categorized as low, moderate, and high using tertiles for easy interpretation of the results.36–38 Furthermore, we linked the geographical covariates data for MDHS 2015–2016 to the children datasets to capture the community malaria prevalence. Malaria prevalence was defined as the average parasite rate of P. falciparum in children between the ages of 2 and 10 years in 2015.39 Community malaria prevalence was categorized as low and high. Communities with malaria prevalence within and below the third quartile40 of overall national communities’ malaria prevalence (15.4%) were labeled as communities with low malaria prevalence, and those with malaria prevalence above the third quartile of overall national communities’ malaria prevalence were labeled as communities with high malaria prevalence.
Statistical analyses.
Descriptive statistics.
Using the “svy” command, the distributions of caregivers’ characteristics according to general and prompt treatment-seeking behaviors were examined using chi-squared tests. The “svy” command considered the clustering and sample weights according to the survey design.
Modeling approaches.
Fixed effects.
A multilevel logistic regression analysis, using the “xtmelogit” command in Stata, was performed to assess the fixed effects of individual- and community-level factors. Four models were fitted in this two-level analysis in which caregivers (Level 1) were nested within their communities (Level 2). Model I was an unconditional model, which only included the outcomes “treatment seeking” and “prompt treatment seeking.” The unconditional model assessed the community effects on the treatment-seeking behavior of the caregivers. Model II fitted the outcome and individual-level variables only, whereas Model III fitted the outcome and community-level variables only. Model IV fitted both the individual- and community-level variables. The fixed effects were reported as adjusted odds ratios (aORs) with their P-values and 95% CIs.
Measures of variation (random effects).
Measures of variation were assessed using several indicators such as area variance (AV) and its 95% CI, the intraclass correlation coefficient (ICC), the median odds ratio (MOR), and the proportional change in variance.
Model fit testing.
The goodness-of-fit of each model was assessed using the Akaike information criterion (AIC), with a lower value representing a closer model fit. The variance inflation factor (VIF) and tolerance were used to assess multicollinearity.41 None of the variables displayed multicollinearity problems (all VIF < 10, tolerance > 0.1).
Interaction between sociodemographics and other key covariates were examined. As none of the interaction terms revealed significant findings (all P-value > 0.1),42 they were excluded from the analysis. In addition, the MDHS asks questions to all women of reproductive age in a sampled household. As such, a single household may have more than one caregiver/child. In total, 100 children (2.4%) came from households with more than one child. A sensitivity analysis to select only one child per household from these households (through a random selection process) was thus conducted, and the results were fairly consistent. Therefore, the findings from the whole sample were reported in this study. All the analyses were performed using Stata version 15.0 (Stata Corp LP, College Station, TX).
Ethical considerations.
All participants provided informed consent to the MDHS surveyors at the start of each individual interview. Ethical approval was obtained from the Malawi National Health Sciences Research Committee of the Malawi Ministry of Health before the survey. Approval to use the MDHS data was provided by the International Classification of Functioning, Disability, and Health under the Demographic Health Surveys (DHS) Program. These datasets are publicly available and can be accessed from the DHS Program on request.
RESULTS
Descriptive statistics.
A total of 4,133 caregivers with febrile children (Level 1) living in 850 communities (Level 2) were analyzed. Overall, 67.3% (2,781) caregivers reported taking their febrile children for treatment. However, only 46.3% (1,914) caregivers reported to have promptly sought treatment. The distribution of characteristics of children’s caregivers according to whether they sought treatment for their febrile child and according to whether they promptly sought treatment for their febrile child is displayed in Table 1. Seeking treatment was positively associated with the child’s age (≥ 24 months), child experiencing cough in the last 2 weeks, child having no diarrhea in the last 2 weeks, having less child, residing in the central or southern region, being a rural dweller, having good health behavior practices, and living in a community with a moderate percentage of rich caregivers (all P < 0.05). The other characteristics were not significantly different. Caregivers with elder children (aged ≥ 24 months), a child who had cough in the last 2 weeks, a child without diarrhea in the last 2 weeks, secondary/postsecondary education, less child, and good health behavior practices and from communities with moderate and high percentages of educated caregivers were positively associated with prompt fever treatment-seeking behavior (all P < 0.05). Living in a community with a moderate or high percentage of caregivers complaining about the distance to a health facility was associated with lower rates of prompt treatment-seeking behavior (P < 0.05).
Table 1.
Bivariate analysis of individual- and community-level factors associated with fever treatment-seeking behaviors
| Variable | Sought treatment for fever | P-value* | Promptly sought treatment | P-value* | ||
|---|---|---|---|---|---|---|
| No (n = 1,352), n (%) | Yes (n = 2,781), n (%) | No (n = 2,219), n (%) | Yes (n = 1,914), n (%) | |||
| Individual-level factors | ||||||
| Gender of child | 0.210 | 0.889 | ||||
| Male | 659 (31.7) | 1,423 (68.3) | 1,120 (53.8) | 962 (46.2) | ||
| Female | 693 (33.8) | 1,358 (66.2) | 1,099 (53.6) | 952 (46.4) | ||
| Child’s age (months) | 0.004 | 0.038 | ||||
| < 6 | 152 (38.2) | 246 (61.8) | 220 (55.3) | 178 (46.2) | ||
| 6 to < 12 | 211 (36.8) | 363 (63.2) | 332 (57.8) | 242 (42.2) | ||
| 12 to < 24 | 390 (34.6) | 736 (65.4) | 630 (56.0) | 496 (44.0) | ||
| ≥ 24 | 599 (29.4) | 1,436 (70.6) | 1,037 (51.0) | 998 (49.0) | ||
| Child had cough | < 0.001 | < 0.001 | ||||
| No | 887 (39.4) | 1,632 (60.6) | 1,982 (57.0) | 967 (43.0) | ||
| Yes | 465 (24.7) | 1,419 (75.3) | 937 (49.7) | 947 (50.3) | ||
| Child had diarrhea | 0.039 | 0.027 | ||||
| No | 810 (31.2) | 1,788 (68.8) | 1,354 (50.1) | 1,244 (47.9) | ||
| Yes | 542 (35.3) | 933 (64.7) | 865 (56.4) | 670 (43.6) | ||
| Caregiver’s age (years) | 0.162 | 0.664 | ||||
| 15–24 | 481 (30.7) | 1,086 (69.3) | 827 (52.8) | 740 (47.2) | ||
| 25–34 | 611 (33.6) | 1,207 (66.4) | 993 (54.6) | 825 (45.4) | ||
| ≥ 35 | 260 (34.8) | 488 (65.2) | 399 (53.3) | 349 (46.7) | ||
| Education | 0.277 | 0.005 | ||||
| No formal education | 189 (36.8) | 325 (63.2) | 309 (60.2) | 204 (39.8) | ||
| Primary | 900 (31.9) | 1,922 (68.1) | 1,523 (54.0) | 1,298 (46.0) | ||
| Secondary/postsecondary | 263 (33.0) | 534 (77.0) | 387 (48.4) | 412 (51.6) | ||
| Occupation | 0.056 | 0.837 | ||||
| Unemployed | 441 (35.3) | 808 (64.7) | 667 (53.4) | 582 (46.6) | ||
| Employed | 911 (31.6) | 1,973 (68.4) | 1,552 (53.8) | 1,332 (46.2) | ||
| Parity | 0.019 | 0.030 | ||||
| One | 301 (29.8) | 708 (70.2) | 519 (51.4) | 490 (48.6) | ||
| Two | 251 (29.9) | 588 (70.1) | 422 (50.3) | 417 (49.7) | ||
| Three+ | 800 (35.0) | 1,485 (65.0) | 1,278 (55.9) | 1,007 (44.1) | ||
| Religion | 0.085 | 0.052 | ||||
| Catholics | 240 (30.8) | 539 (69.2) | 401 (51.5) | 378 (48.5) | ||
| Protestants | 245 (29.3) | 591 (70.7) | 418 (50.0) | 418 (50.0) | ||
| Muslims and others | 867 (34.4) | 1,651 (65.6) | 1,400 (55.6) | 1,118 (44.4) | ||
| Marital status | 0.788 | 0.840 | ||||
| Unmarried | 235 (33.3) | 471 (66.7) | 382 (54.1) | 324 (45.9) | ||
| Married | 1,117 (32.6) | 2,310 (67.4) | 1,837 (53.6) | 1,590 (46.4) | ||
| Number of under-five children | 0.946 | 0.956 | ||||
| One | 796 (32.5) | 1,652 (67.5) | 1,312 (53.6) | 1,136 (46.4) | ||
| Two | 464 (32.8) | 949 (67.2) | 758 (53.6) | 655 (46.4) | ||
| Three+ | 92 (33.8) | 180 (66.2) | 149 (54.8) | 123 (45.2) | ||
| Media exposure | 0.863 | 0.951 | ||||
| No | 863 (32.6) | 1,783 (67.4) | 1,419 (53.6) | 1,227 (46.4) | ||
| Yes | 489 (32.9) | 998 (67.1) | 780 (48.8) | 787 (51.2) | ||
| Health behavior | < 0.001 | < 0.001 | ||||
| Poor | 208 (44.1) | 264 (55.9) | 305 (64.6) | 167 (35.4) | ||
| Moderate | 682 (33.3) | 1,367 (66.7) | 1,114 (54.4) | 935 (45.6) | ||
| Good | 462 (28.7) | 1,150 (71.3) | 800 (49.6) | 812 (50.4) | ||
| Distance to a health facility | 0.691 | 0.743 | ||||
| No problem | 527 (33.2) | 1,060 (66.8) | 846 (53.3) | 741 (46.7) | ||
| Problem | 825 (32.4) | 1,721 (67.6) | 1,373 (53.9) | 1,173 (46.1) | ||
| Region | 0.018 | 0.219 | ||||
| Northern | 183 (41.0) | 263 (51.0) | 262 (58.8) | 184 (41.2) | ||
| Central | 577 (31.8) | 1,239 (68.2) | 960 (52.9) | 856 (47.1) | ||
| Southern | 592 (31.6) | 1,279 (68.4) | 997 (53.3) | 874 (46.7) | ||
| Residence | 0.003 | 0.958 | ||||
| Urban | 188 (41.2) | 268 (58.8) | 244 (53.6) | 211 (46.4) | ||
| Rural | 1,164 (31.7) | 2,513 (68.3) | 1,975 (53.7) | 1,703 (46.3) | ||
| Wealth | 0.318 | 0.618 | ||||
| Poor | 629 (31.8) | 1,352 (68.2) | 1,075 (54.3) | 906 (45.7) | ||
| Middle | 269 (31.7) | 580 (68.3) | 643 (62.5) | 386 (37.5) | ||
| Rich | 454 (34.8) | 849 (65.2) | 681 (52.3) | 622 (47.7) | ||
| Gender of head of household | 0.113 | 0.805 | ||||
| Male | 1,019 (33.6) | 2,015 (66.4) | 1,624 (53.5) | 1,410 (46.5) | ||
| Female | 333 (30.3) | 766 (69.7) | 595 (54.1) | 504 (45.9) | ||
| Community-level factors | ||||||
| Community malaria prevalence | 0.694 | 0.475 | ||||
| Low | 991 (32.5) | 2,059 (67.5) | 1,623 (53.2) | 1,427 (46.8) | ||
| High | 361 (33.3) | 722 (66.7) | 596 (55.0) | 487 (45.0) | ||
| Community employment | 0.094 | 0.194 | ||||
| Low | 455 (35.9) | 811 (64.1) | 699 (55.2) | 568 (44.8) | ||
| Moderate | 355 (30.3) | 817 (69.7) | 595 (50.8) | 577 (49.2) | ||
| High | 542 (32.0) | 1,153 (68.0) | 925 (54.6) | 769 (45.4) | ||
| Community education | 0.092 | 0.011 | ||||
| Low | 578 (35.3) | 1,060 (64.7) | 943 (57.5) | 696 (42.5) | ||
| Moderate | 389 (30.0) | 906 (70.0) | 667 (51.5) | 628 (48.5) | ||
| High | 385 (32.1) | 815 (67.9) | 609 (50.8) | 590 (49.2) | ||
| Community wealth | 0.001 | 0.129 | ||||
| Low | 574 (33.2) | 1,155 (66.8) | 965 (55.8) | 764 (44.2) | ||
| Moderate | 398 (28.1) | 1,018 (71.9) | 722 (51.0) | 694 (58.8) | ||
| High | 380 (38.5) | 608 (61.5) | 532 (53.8) | 456 (46.2) | ||
| Community distance to a health facility | 0.333 | 0.024 | ||||
| Low | 315 (32.2) | 662 (67.8) | 476 (48.7) | 502 (51.3) | ||
| Moderate | 467 (31.0) | 1,038 (79.0) | 817 (54.3) | 688 (45.7) | ||
| High | 570 (34.5) | 1,081 (65.5) | 926 (56.1) | 724 (43.9) | ||
Bold = significant (P < 0.05).
* Pearson’s χ2 test.
Modeling approaches.
Fixed effects.
Table 2 reveals the fixed effects of individual- and community-level factors on fever treatment-seeking behavior among caregivers. In Model IV, febrile children aged 6–11, 12–23, and ≥ 24 months were more likely to be taken for treatment than those younger than 6 months (aOR = 1.50, 95% CI = 1.11–2.03; aOR = 1.49, 95% CI = 1.14–1.96; and aOR = 1.92, 95% CI = 1.48–2.49, respectively). Children who had a cough within 2 weeks of the survey were more likely to be brought for treatment (aOR = 2.12, 95% CI = 1.82–2.47) than those who had no cough. Model IV further revealed that children who had diarrhea within 2 weeks before the survey, and whose caregivers had three or more children and belonged to the Muslims and other religious groups were less likely to be taken for treatment (aOR = 0.80, 95% CI = 0.68–0.93; aOR = 0.69, 95% CI = 0.51–0.92; and aOR = 0.70, 95% CI = 0.57–0.86, respectively), compared with their respective reference groups. Children from the central and southern regions (aOR = 1.95, 95% CI = 1.46–2.61 and aOR = 1.67, 95% CI = 1.27–2.20, respectively) were more likely to be taken for treatment than those from the northern region. Caregivers with moderate and good health behaviors were more likely to seek treatment for their febrile children than those with poor health behaviors (aOR = 1.53, 95% CI = 1.21–1.93 and aOR = 1.81, 95% CI = 1.41–2.32, respectively). Children living in communities with a high prevalence of malaria were more likely to be taken for treatment than those living in communities with a low prevalence of malaria (aOR = 1.26, 95% CI = 1.01–1.59). Children from communities with moderate and high percentages of educated caregivers were more likely to be taken for treatment (aOR = 1.26, 95% CI = 1.01–1.58 and aOR = 1.31, 95% CI = 1.02–1.70, respectively) than those from communities with low education. Children from communities with moderate and high percentages of caregivers perceiving the distance to a health facility as a problem were less likely to be taken for treatment (aOR = 0.74, 95% CI = 0.58–0.96 and aOR = 0.67, 95% CI = 0.51–0.88, respectively) than those from communities with a low percentage of caregivers perceiving the distance to a health facility as a problem.
Table 2.
Measures of association between individual- and community-level factors and fever treatment-seeking behavior
| Variable | Model I | Model II | Model III | Model IV | ||||
|---|---|---|---|---|---|---|---|---|
| aOR (95% CI) | P-value | aOR (95% CI) | P-value | aOR (95% CI) | P-value | aOR (95% CI) | P-value | |
| Individual-level factor | ||||||||
| Gender of child | ||||||||
| Male | – | – | 1.00 | – | – | 1.00 | ||
| Female | – | – | 0.93 (0.81–1.08) | 0.393 | – | – | 0.93 (0.81–1.08) | 0.360 |
| Child’s age (months) | ||||||||
| < 6 | – | – | 1.00 | – | – | 1.00 | – | |
| 6 to < 12 | – | – | 1.49 (1.10–2.02) | 0.010 | – | – | 1.50 (1.11–2.03) | 0.008 |
| 12 to < 24 | – | – | 1.49 (1.13–1.95) | 0.004 | – | – | 1.49 (1.14–1.96) | 0.004 |
| ≥ 24 | – | – | 1.92 (1.36–2.29) | < 0.001 | – | – | 1.92 (1.48–2.49) | < 0.001 |
| Child had cough | ||||||||
| No | – | – | 1.00 | – | – | 1.00 | ||
| Yes | – | – | 2.14 (1.83–2.50) | < 0.001 | – | – | 2.12 (1.82–2.47) | < 0.001 |
| Child had diarrhea | ||||||||
| No | – | – | 1.00 | – | – | 1.00 | ||
| Yes | – | – | 0.79 (0.67–0.93) | 0.004 | – | – | 0.80 (0.68–0.93) | 0.005 |
| Caregiver’s age (years) | ||||||||
| 15–24 | – | – | 1.00 | – | – | 1.00 | ||
| 25–34 | – | – | 1.06 (0.84–1.35) | 0.627 | – | – | 1.05 (0.82–1.33) | 0.700 |
| ≥ 35 | – | – | 0.97 (0.72–1.31) | 0.845 | – | – | 0.97 (0.72–1.31) | 0.835 |
| Education | ||||||||
| No formal education | – | – | 1.00 | – | – | 1.00 | ||
| Primary | – | – | 1.16 (0.91–1.47) | 0.223 | – | – | 1.05 (0.82–1.34) | 0.697 |
| Secondary/post-secondary | – | – | 0.96 (0.71–1.30) | 0.795 | – | – | 0.87 (0.64–1.18) | 0.368 |
| Employment | ||||||||
| Unemployed | – | – | 1.00 | – | – | 1.00 | ||
| Employed | – | – | 1.16 (0.98–1.37) | 0.079 | – | – | 1.12 (0.94–1.33) | 0.214 |
| Parity | ||||||||
| One | – | – | 1.00 | – | – | 1.00 | ||
| Two | – | – | 0.80 (0.63–1.03) | 0.087 | – | – | 0.81 (0.63–1.04) | 0.099 |
| Three+ | – | – | 0.68 (0.51–0.92) | 0.011 | – | – | 0.69 (0.51–0.92) | 0.013 |
| Religion | ||||||||
| Catholics | – | – | 1.00 | – | – | 1.00 | ||
| Protestants | – | – | 0.90 (0.70–1.15) | 0.379 | – | – | 0.89 (0.69–1.14) | 0.347 |
| Muslims and others | – | – | 0.70 (0.57–0.85) | < 0.001 | – | – | 0.70 (0.57–0.86) | 0.001 |
| Marital status | ||||||||
| Unmarried | – | – | 1.00 | – | – | 1.00 | ||
| Married | – | – | 1.10 (0.88–1.38) | 0.416 | – | – | 1.11 (0.88–1.39) | 0.383 |
| Number of under-five children | ||||||||
| One | – | – | 1.00 | – | – | 1.00 | ||
| Two | – | – | 1.19 (0.99–1.42) | 0.053 | – | – | 1.20 (1.00–1.43) | 0.077 |
| Three+ | – | – | 1.29 (0.94–1.76) | 0.112 | – | – | 1.32 (0.97–1.82) | 0.078 |
| Media exposure | ||||||||
| No | – | – | 1.00 | – | – | 1.00 | ||
| Yes | – | – | 1.09 (0.93–1.29) | 0.293 | – | – | 1.10 (0.93–1.29) | 0.263 |
| Health behavior | ||||||||
| Poor | – | – | 1.00 | – | – | 1.00 | ||
| Moderate | – | – | 1.53 (1.21–1.93) | < 0.001 | – | – | 1.53 (1.21–1.93) | < 0.001 |
| Good | – | – | 1.87 (1.46–2.40) | < 0.001 | – | – | 1.81 (1.41–2.32) | < 0.001 |
| Distance to a health facility | ||||||||
| Problem | – | – | 1.00 | – | – | 1.00 | ||
| No problem | – | – | 1.09 (0.93–1.28) | 0.290 | – | – | 0.99 (0.83–1.18) | 0.945 |
| Region | ||||||||
| Northern | – | – | 1.00 | – | – | 1.00 | ||
| Central | – | – | 1.78 (1.37–2.30) | < 0.001 | – | – | 1.95 (1.46–2.61) | < 0.001 |
| Southern | – | – | 1.77 (1.38–2.26) | < 0.001 | – | – | 1.67 (1.27–2.20) | < 0.001 |
| Residence | ||||||||
| Urban | – | – | 1.00 | – | – | 1.00 | ||
| Rural | – | – | 0.98 (0.74–1.28) | 0.871 | – | – | 1.04 (0.75–1.44) | 0.803 |
| Wealth | ||||||||
| Poor | 1.00 | 1.00 | ||||||
| Middle | – | – | 0.99 (0.81–1.21) | 0.919 | – | – | 0.96 (0.79–1.18) | 0.731 |
| Rich | – | – | 0.98 (0.81–1.21) | 0.952 | – | – | 0.98 (0.80–1.22) | 0.876 |
| Gender of head of household | ||||||||
| Male | – | – | 1.00 | – | – | 1.00 | ||
| Female | – | – | 1.06 (0.87–1.30) | 0.558 | – | – | 1.06 (0.87–1.30) | 0.552 |
| Community-level factors | ||||||||
| Community malaria prevalence | ||||||||
| Low | – | – | – | – | 1.00 | 1.00 | ||
| High | – | – | – | – | 1.20 (0.99–1.45) | 0.053 | 1.26 (1.01–1.59) | 0.048 |
| Community employment | ||||||||
| Low | – | – | – | – | 1.00 | 1.00 | ||
| Moderate | – | – | – | – | 1.31 (1.06–1.61) | 0.012 | 1.21 (0.97–1.50) | 0.090 |
| High | – | – | – | – | 1.25 (1.02–1.54) | 0.032 | 1.05 (0.82–1.32) | 0.693 |
| Community education | ||||||||
| Low | – | – | – | – | 1.00 | 1.00 | ||
| Moderate | – | – | – | – | 1.34 (1.08–1.66) | 0.007 | 1.26 (1.01–1.58) | 0.045 |
| High | – | – | – | – | 1.29 (1.03–1.64) | 0.026 | 1.31 (1.02–1.70) | 0.037 |
| Community wealth | ||||||||
| Low | – | – | – | – | 1.00 | 1.00 | ||
| Moderate | – | – | – | – | 1.21 (0.98–1.49) | 0.072 | 1.24 (0.99–1.54) | 0.053 |
| High | – | – | – | – | 0.75 (0.58–0.97) | 0.027 | 0.80 (0.59–1.09) | 0.162 |
| Community distance to a health facility | ||||||||
| Low | – | – | – | – | 1.00 | 1.00 | ||
| Moderate | – | – | – | – | 0.77 (0.61–0.96) | 0.022 | 0.74 (0.58–0.96) | 0.021 |
| High | – | – | – | – | 0.65 (0.51–0.83) | < 0.001 | 0.67 (0.51–0.88) | 0.005 |
| Measures of variation | ||||||||
| Random effects | ||||||||
| Variance (95% CI) | 0.39 (0.26–0.59) | < 0.001 | 0.37 (0.24–0.58) | < 0.001 | 0.32 (0.20–0.51) | < 0.001 | 0.34 (0.22–0.54) | < 0.001 |
| ICC (%) | 10.6 | 10.1 | 8.9 | 9.4 | ||||
| PCV | Ref | 5.1 | 18.0 | 12.8 | ||||
| MOR | 1.81 | 1.79 | 1.72 | 1.75 | ||||
| Model fit | ||||||||
| AIC | 4,921.7 | 4,773.5 | 4,895.9 | 4,760.8 | ||||
AIC = Akaike information criterion; aOR = adjusted odds ratio; ICC = intraclass correlation coefficient; MOR = median odds ratio; PVC = proportional change in variance.
Bold = significant (P < 0.05).
Model I: unconditional model with random intercepts and no predictors.
Model II: adjusted for individual-level factors only: gender of the child and child’s age; caregiver’s age, education, employment, parity, religion, and marital status; number of under-five children of the caregiver; and caregiver’s exposure to media, region, residence, wealth, perception about the distance to a health facility, and health behavior.
Model III: adjusted for community-level factors: community employment, community education, community wealth, and community distance to a health facility.
Model IV: controlled for both individual- and community-level factors in Models II and III.
The fixed effects of individual- and community-level factors influencing prompt treatment-seeking behaviors are displayed in Table 3. In Model IV, children who had diarrhea within 2 weeks before the survey and whose caregivers belonged to the Muslim and other religious groups had reduced odds of being promptly taken for treatment (aOR = 0.85, 95% CI = 0.74–0.98 and aOR = 0.78, 95% CI = 0.65–0.93, respectively), compared with those who had no diarrhea and whose caregivers belonged to the Catholic group, respectively. Children who had a cough within 2 weeks before the survey and from the central and southern regions were more likely to be promptly taken for treatment (aOR = 1.24, 95% CI = 1.08–1.42; aOR = 1.70, 95% CI = 1.30–2.21; and aOR = 1.48, 95% CI = 1.15–1.90, respectively) than their respective reference groups. Caregivers with moderate and good health behaviors were more likely to promptly seek treatment for their febrile children than those with poor health behaviors (aOR = 1.56, 95% CI = 1.24–1.96 and aOR = 1.83, 95% CI = 1.44–2.32, respectively). The community-level factors revealed that children from communities with moderate and high percentages of educated caregivers had increased odds of being promptly taken for treatment (aOR = 1.26, 95% CI = 1.03–1.55 and aOR = 1.27, 95% CI = 1.01–1.59, respectively) compared with those from communities with a low percentage of educated caregivers. However, children from communities with moderate and high percentages of caregivers complaining about the distance to a health facility were less likely to be promptly taken for treatment (aOR = 0.69, 95% CI = 0.55–0.86 and aOR = 0.65, 95% CI = 0.51–0.84, respectively) than those from communities with a low percentage of caregivers complaining about the distance to a health facility.
Table 3.
Measures of association between individual- and community-level factors and promptness in fever treatment-seeking behavior
| Variable | Model I | Model II | Model III | Model IV | ||||
|---|---|---|---|---|---|---|---|---|
| aOR (95% CI) | P-value | aOR (95% CI) | P-value | aOR (95% CI) | P-value | aOR (95% CI) | P-value | |
| Individual-level factor | ||||||||
| Gender of the child | ||||||||
| Male | – | – | 1.00 | – | – | 1.00 | ||
| Female | – | – | 0.98 (0.86–1.12) | 0.735 | – | – | 0.98 (0.85–1.11) | 0.713 |
| Child’s age (months) | ||||||||
| < 6 | – | – | 1.00 | – | – | 1.00 | ||
| 6 to < 12 | – | – | 1.03 (0.78–1.37) | 0.822 | – | – | 1.04 (0.78–1.38) | 0.792 |
| 12 to < 24 | – | – | 0.99 (0.77–1.28) | 0.953 | – | – | 0.99 (0.77–1.29) | 0.997 |
| ≥ 24 | – | – | 1.25 (0.98–1.60) | 0.068 | – | – | 1.25 (0.98–1.60) | 0.069 |
| Child had cough | ||||||||
| No | – | – | 1.00 | – | – | 1.00 | ||
| Yes | – | – | 1.25 (1.09–1.44) | 0.001 | – | – | 1.24 (1.08–1.42) | 0.002 |
| Child had diarrhea | ||||||||
| No | – | – | 1.00 | – | – | 1.00 | ||
| Yes | – | – | 0.85 (0.73–0.98) | 0.026 | – | – | 0.85 (0.74–0.98) | 0.030 |
| Caregiver’s age (years) | ||||||||
| 15–24 | – | – | 1.00 | – | – | 1.00 | ||
| 25–34 | – | – | 1.02 (0.82–1.27) | 0.843 | – | – | 1.01 (0.81–1.25) | 0.946 |
| ≥ 35 | – | – | 1.09 (0.83–1.44) | 0.537 | – | – | 1.09 (0.82–1.43) | 0.558 |
| Education | ||||||||
| No formal education | – | – | 1.00 | – | – | 1.00 | ||
| Primary | – | – | 1.19 (0.95–1.49) | 0.123 | – | – | 1.09 (0.87–1.38) | 0.443 |
| Secondary/postsecondary | – | – | 1.30 (0.98–1.72) | 0.070 | – | – | 1.18 (0.89–1.57) | 0.256 |
| Employment | ||||||||
| Unemployed | – | – | 1.00 | – | – | 1.00 | ||
| Employed | – | – | 0.94 (0.81–1.09) | 0.434 | – | – | 0.95 (0.81–1.11) | 0.499 |
| Parity | ||||||||
| One | – | – | 1.00 | – | – | 1.00 | ||
| Two | – | – | 0.95 (0.76–1.18) | 0.636 | – | – | 0.95 (0.76–1.19) | 0.654 |
| Three+ | – | – | 0.85 (0.65–1.10) | 0.218 | – | – | 0.85 (0.65–1.11) | 0.234 |
| Religion | ||||||||
| Catholics | – | – | 1.00 | – | – | 1.00 | ||
| Protestants | – | – | 0.97 (0.78–1.20) | 0.775 | – | – | 0.96 (0.77–1.19) | 0.698 |
| Muslims and others | – | – | 0.78 (0.65–0.93) | 0.006 | – | – | 0.78 (0.65–0.93) | 0.007 |
| Marital status | ||||||||
| Unmarried | – | – | 1.00 | – | – | 1.00 | ||
| Married | – | – | 0.89 (0.73–1.10) | 0.284 | – | – | 0.89 (0.73–1.10) | 0.278 |
| Number of under-five children | ||||||||
| One | – | – | 1.00 | – | – | 1.00 | ||
| Two | – | – | 1.15 (0.98–1.35) | 0.090 | – | – | 1.16 (0.98–1.36) | 0.076 |
| Three+ | – | – | 1.24 (0.93–1.65) | 0.145 | – | – | 1.26 (0.95–1.69) | 0.109 |
| Media exposure | ||||||||
| No | – | – | 1.00 | – | – | 1.00 | ||
| Yes | – | – | 0.95 (0.82–1.11) | 0.536 | – | – | 0.96 (0.82–1.11) | 0.551 |
| Health behavior | ||||||||
| Poor | – | – | 1.00 | – | – | 1.00 | ||
| Moderate | – | – | 1.56 (1.24–1.96) | < 0.001 | – | – | 1.56 (1.24–1.96) | < 0.001 |
| Good | – | – | 1.85 (1.46–2.35) | < 0.001 | – | – | 1.83 (1.44–2.32) | < 0.001 |
| Distance to a health facility | ||||||||
| Problem | – | – | 1.00 | – | – | 1.00 | ||
| No problem | – | – | 1.06 (0.92–1.22) | 0.440 | – | – | 0.95 (0.81–1.12) | 0.555 |
| Region | ||||||||
| Northern | – | – | 1.00 | – | – | 1.00 | ||
| Central | – | – | 1.49 (1.17–1.89) | 0.001 | – | – | 1.70 (1.30–2.21) | < 0.001 |
| Southern | – | – | 1.47 (1.17–1.84) | 0.001 | – | – | 1.48 (1.15–1.90) | 0.002 |
| Residence | ||||||||
| Urban | – | – | 1.00 | – | – | 1.00 | ||
| Rural | – | – | 0.85 (0.66–1.09) | 0.190 | – | – | 1.01 (0.75–1.35) | 0.960 |
| Wealth | ||||||||
| Poor | – | – | 1.00 | – | – | 1.00 | ||
| Middle | – | – | 1.01 (0.84–1.21) | 0.920 | – | – | 0.98 (0.81–1.18) | 0.820 |
| Rich | – | – | 1.14 (0.95–1.37) | 0.152 | – | – | 1.11 (0.91–1.34) | 0.295 |
| Gender of head of household | ||||||||
| Male | – | – | 1.00 | – | – | 1.00 | ||
| Female | – | – | 0.87 (0.73–1.04) | 0.134 | – | – | 0.87 (0.73–1.04) | 0.134 |
| Community-level factors | ||||||||
| Community malaria prevalence | ||||||||
| Low | – | – | – | – | 1.00 | 1.00 | ||
| High | – | – | – | – | 1.11 (0.93–1.32) | 0.259 | 1.17 (0.95–1.45) | 0.142 |
| Community employment | ||||||||
| Low | – | – | – | – | 1.00 | 1.00 | ||
| Moderate | – | – | – | – | 1.11 (0.92–1.35) | 0.281 | 1.08 (0.88–1.31) | 0.475 |
| High | – | – | – | – | 1.00 (0.82–1.21) | 0.996 | 0.91 (0.73–1.13) | 0.387 |
| Community education | ||||||||
| Low | – | – | – | – | 1.00 | 1.00 | ||
| Moderate | – | – | – | – | 1.33 (1.09–1.61) | 0.006 | 1.26 (1.03–1.55) | 0.025 |
| High | – | – | – | – | 1.25 (1.01–1.55) | 0.041 | 1.27 (1.01–1.59) | 0.045 |
| Community wealth | ||||||||
| Low | – | – | – | – | 1.00 | 1.00 | ||
| Moderate | – | – | – | – | 1.14 (0.94–1.37) | 0.190 | 1.13 (0.93–1.38) | 0.219 |
| High | – | – | – | – | 0.91 (0.72–1.16) | 0.455 | 0.89 (0.67–1.18) | 0.429 |
| Community distance to a health facility | ||||||||
| Low | – | – | – | – | 1.00 | 1.00 | ||
| Moderate | – | – | – | – | 0.69 (0.56–0.85) | < 0.001 | 0.69 (0.55–0.86) | 0.001 |
| High | – | – | – | – | 0.64 (0.51–0.79) | < 0.001 | 0.65 (0.51–0.84) | 0.001 |
| Measures of variation | ||||||||
| Random effects | ||||||||
| Variance (95% CI) | 0.36 (0.25–0.53) | < 0.001 | 0.33 (0.22–0.49) | < 0.001 | 0.31 (0.20–0.47) | < 0.001 | 0.29 (0.19–0.46) | < 0.001 |
| ICC (%) | 9.9 | 9.1 | 8.6 | 8.1 | ||||
| PCV | Ref | 8.3 | 13.9 | 19.4 | ||||
| MOR | 1.77 | 1.73 | 1.70 | 1.68 | ||||
| Model fit | ||||||||
| AIC | 5,565.7 | 5,524.9 | 5,545.6 | 5,515.0 | ||||
AIC = Akaike information criterion; aOR = adjusted odds ratio; ICC = intraclass correlation coefficient; MOR = median odds ratio; PVC = proportional change in variance.
Bold = significant (P < 0.05).
Model I: unconditional model with random intercepts and no predictors.
Model II: adjusted for individual-level factors only: gender of the child and child’s age; caregiver’s age, education, employment, parity, religion, and marital status; number of under-five children of the caregiver; and caregiver’s exposure to media, region, residence, wealth, perception about the distance to a health facility, and health behavior.
Model III: adjusted for community-level factors: community employment, community education, community wealth, and community distance to a health facility.
Model IV: controlled for both individual- and community-level factors in Models II and III.
Measures of variation (random effects).
Table 2 further presents the random effects of communities on treatment-seeking behaviors. Results from Model I revealed that a significant variation was observed in treatment-seeking behaviors across the communities (AV = 0.39, 95% CI = 0.26–0.59). The variation decreased but remained significant even after adjusting for individual- and community-level factors in Model IV (AV = 0.34, 95% CI = 0.22–0.54). The ICC in Model I demonstrated that 10.6% of the total variation was due to unobserved community effects; after adjustments in Model IV, 9.4% of the total variation was due to unobserved community characteristics. Model IV further revealed that the MOR showing the effects of community heterogeneity was 1.75. The individual- and community-level factors in the analysis explained 12.8% of the total variation in treatment-seeking behavior. Last, the AIC showed that Model IV was a better model (i.e., lower AIC than of other models). Similar results were observed for prompt treatment-seeking behaviors (Table 3).
DISCUSSION
This is the first population-based study that examined both individual- and community-level factors associated with fever treatment-seeking behaviors in Malawi. The results demonstrated that treatment-seeking behavior of caregivers with febrile children aged 5 years and younger is influenced by socioeconomic and health-related factors operating at both individual and community levels in Malawi. Although few studies have examined fever treatment-seeking behaviors in Malawi,43,44 the present study results revealed significant community effects and compared general and prompt fever treatment-seeking behaviors. Specifically, febrile children living in communities with high malaria prevalence and from communities with moderate and high percentages of educated caregivers were more likely to be taken for treatment than those living in communities with low malaria prevalence and a low percentage of educated caregivers, respectively. Caregivers with febrile children from the central and southern regions exhibited higher odds of seeking treatment than their northern region counterparts.
The prevalence of treatment-seeking behavior among caregivers with febrile children was 67.3%. However, only 46.3% promptly sought treatment. A study in Liberia reported that 98.5% of caregivers with children having fever in the last 2 weeks had sought treatment.25 The higher prevalence could be attributed to the fact that the Liberian study was conducted following a social and behavior change campaign, and probably, most women had just been exposed to malaria-related messages. In terms of prompt treatment-seeking behavior, a similar lower rate of 42% was observed in a longitudinal study in Zambia.22 Previous studies have reported that home-based fever treatment practice may be among the reasons for delayed treatment-seeking behaviors.16,29 The lower prevalence of timely treatment-seeking behaviors highlights the importance of formulating strategies to improve caregivers’ ability to recognize and seek appropriate care for fever.
Regional variation in terms of seeking treatment and promptness in seeking treatment was observed in Malawi. Similarly, in Liberia, regional differences were observed in treatment-seeking behaviors.25 A study in four rural and urban health centers in Malawi revealed that 25% of all fevers in children were malaria related.45 The Malawi National Malaria Control Programme policy then recommends prompt parasitological confirmation among fever cases before treatment commences.21 As a result, febrile case management is integrated with malaria programs in Malawi.17 A higher prevalence of malaria has been reported in the southern and central regions of Malawi than in the northern region.46 Subsequently, malaria programs, messages, and efforts may be more concentrated in these high-prevalence regions, making the people in these regions more aware of how they should act when their children have fever. Febrile children whose caregivers belonged to the Muslim and other religious groups were less likely to be taken for treatment and promptly taken for treatment. Religious beliefs have been associated with the use of health services.47–49 In Ghana, religious beliefs are associated with adherence to medication.50 Therefore, qualitative studies are necessary for Malawi to understand different religious beliefs and their influence on treatment-seeking behaviors.
A higher parity was associated with reduced odds of seeking treatment for febrile children. This may be because caregivers with more children may have more responsibility to appropriately take care of their children than those with fewer children.10,51 In addition, multiparous caregivers are more likely to be older and may believe that they have the necessary skills based on their previous experiences to take care of the children.30 Surprisingly, children older than 6 months were more likely to be brought in for treatment than those younger than 6 months. Most previous studies have reported that caregivers with infants are more likely to seek treatment than those with older children.28 The present study results are consistent with those of an Indian study, where prevalence of treatment seeking for younger children was lower than those for older children.52 Younger children may experience frequent illnesses, most of which get resolved on their own.52 Nevertheless, future studies should explore further about the association between a child’s age and fever treatment-seeking behaviors in Malawi.
Children who were reported to have had a cough in the last 2 weeks were more likely to be brought for fever treatment than those who had no cough. Children with cough may produce other symptoms such as difficult breathing53 which may drive caregivers to seek care. Febrile children who had diarrhea in the last 2 weeks were less likely to be brought for treatment than those who had no diarrhea. Women are mostly taught on how to manage diarrhea cases at home using oral rehydration therapy (ORT).54 A 2016 study in southern Malawi revealed that a higher proportion (68%) of women with children with diarrhea reported administering ORT at home, whereas 14% gave the children other fluids and only 13% visited a health facility.55 In addition, women may regard diarrhea as a minor ailment56 which can be easily managed at home without incurring further costs by visiting health facilities. Prior health service utilization was associated with increased odds of both treatment-seeking behaviors. This may indicate that caregivers with a history of more health service utilization probably possess better health behaviors, including seeking treatment for their febrile younger ones. In India, women who had fewer ANC visits were less likely to seek treatment for their sick children.52
Community characteristics have been reported to shape and influence different health outcomes.57 Individuals living in the same communities are exposed to a common set of social climate that may influence decision-making and health behaviors.58 Living in communities with high prevalence of malaria was associated with increased odds of treatment-seeking behavior for febrile children among caregivers. Because of high malaria prevalence, caregivers may be more likely to regard every fever case to be malaria related and, thus, rush to seek treatment at a health facility. In addition, women in these communities are more likely to have more information about malaria, as malaria health education programs may focus their efforts in these areas compared with low-prevalence areas. In Uganda, the odds of seeking treatment increased when caregivers had knowledge about malaria.59 The finding that caregivers from communities with moderate and high percentages of educated caregivers were more likely to seek treatment and promptly seek treatment emphasizes the importance of women education and is similar to the results reported in Tanzania and Ethiopia.26,60,61 Educated caregivers are more likely to comprehend health-care messages.62,63 A study in Britain revealed that education may also enhance demand for services.64 Women who are educated are also more likely to have access to information than women without formal education.65 Once a community has a high percentage of these women, they are more likely to influence the decisions of fellow women by sharing their knowledge with others within the community, in this case, about fever and appropriate treatment seeking.66 Women’s education is a form of empowerment that subsequently helps women to make independent decisions, and this has been reported to influence the uptake of health services, which might ultimately include appropriate treatment-seeking behaviors.32,64,67
The analysis revealed that the perceived distance to a health facility at an individual level was not significantly associated with treatment-seeking behaviors. This is in contrast to the results of studies conducted in Mozambique, Liberia, and Thailand–Myanmar border.68–70 The disparities observed in this study and previous literature could be explained by how the information on distance to a health facility was captured. The present study collected information on caregivers’ perception on the distance to the nearest health facility, which may vary because of caregivers’ context, history, and beliefs. The Mozambican study measured distance according to the time (in hours) taken from households to reach the nearest health facility.68 However, at the community level, children from communities with moderate and high percentages of caregivers who perceived the distance to a health facility as a problem were less likely to be taken for treatment and less likely to be promptly taken for treatment. Another Liberian study speculated that unmeasured community effects on treatment-seeking behaviors from their study could be due to the distance to a health facility.25 Although individual-level factors may be stronger predictors of health outcomes than community-level factors, the study results indicated that improving community access to health facilities is of paramount importance.71 However, the results in the present study should be interpreted with caution as perceived but not actual distance to a health facility was assessed.
This study used nationally representative samples, and, therefore, results could be generalized to all Malawian caregivers. In addition, although the information on the history of fever was based on recall, the 2-week recall period could help reduce recall bias. However, this study had some limitations, too. First, as a cross-sectional study, causal inferences could not be drawn. Second, the use of clusters or administratively defined boundaries could yield information bias for those caregivers from unfitted administrative communities. Third, self-reports on whether treatment was sought are prone to social desirability bias. Caregivers may attempt to present that they take good care of their children by, among others, seeking treatment whenever their children have fever. Non-differential misclassification may have biased the association toward the null. Fourth, the study was not able to examine the influence of caregivers’ knowledge, perceptions, and attitudes on different childhood illnesses that provoke fever such as malaria, which may play a role in influencing whether one seeks treatment and could help explain the significant unmeasured community effects observed in this study.
CONCLUSION
The study revealed that 67.3% of the caregivers reported seeking treatment for their febrile child, whereas 46.3% reported promptly seeking treatment. Therefore, in Malawi, improving prompt treatment-seeking behavior should be emphasized. In an effort to improve the management of childhood illnesses, including the life-threatening disease of malaria, public health practitioners should consider that a low community education level, evident regional variations, and perceived community distance to health facilities may derail treatment-seeking behaviors. Programs that consider both individual and community characteristics may prove vital to effect improved behaviors. Hence, the results form a basis for future programs to conduct proper profiling of communities when implementing programs focusing on febrile case management. In addition, messages on treatment-seeking behaviors for febrile children should focus more on multiparous women. Future studies should explore other institution-based factors, such as availability of skilled personnel and drugs, which may influence treatment-seeking behaviors of caregivers.
Acknowledgments:
We would like to acknowledge the International Classification of Functioning, Disability, and Health for providing permission to use the MDHS dataset for this analysis.
REFERENCES
- 1.Barbi E, Marzuillo P, Neri E, Naviglio S, Krauss BS, 2017. Fever in children: pearls and pitfalls. Children 4: 81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Olotu A, Fegan G, Williams TN, Sasi P, Ogada E, Bauni E, Wambua J, Marsh K, Borrmann S, Bejon P, 2010. Defining clinical malaria: the specificity and incidence of endpoints from active and passive surveillance of children in rural Kenya. PLoS One 5: e15569. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Stauffer W, Fischer PR, 2003. Diagnosis and treatment of malaria in children. Clin Infect Dis 37: 1340–1348. [DOI] [PubMed] [Google Scholar]
- 4.Schumacher R-F, Spinelli E, 2012. Malaria in children. Mediterr J Hematol Infect Dis 4: e2012073. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Tarimo DS, Lyimo EO, Moshiro C, 2014. Accuracy of fever and fraction of fevers attributable to malaria among under-fives under reduced malaria infection prevalence in district. Malar Chemother Control Elimin 3: 1. [Google Scholar]
- 6.WHO , 2015. Guidelines for the Treatment of Malaria, 3rd edition Geneva, Switzerland: World Health Organization. [Google Scholar]
- 7.Gething PW, Kirui VC, Alegana VA, Okiro EA, Noor AM, Snow RW, 2010. Estimating the number of paediatric fevers associated with malaria infection presenting to Africa’s public health sector in 2007. PLoS Med 7: e1000301. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Okiro EA, Snow RW, 2010. The relationship between reported fever and Plasmodium falciparum infection in African children. Malar J 9: 99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.WHO , 2017. World Malaria Report 2017. Geneva, Switzerland: World Health Organization. [Google Scholar]
- 10.Mitiku I, Assefa A, 2017. Caregivers’ perception of malaria and treatment-seeking behaviour for under five children in Mandura district, west Ethiopia: a cross-sectional study. Malar J 16: 144. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Salah MT, Adam I, Malik EM, 2007. Care-seeking behavior for fever in children under five years in an urban area in eastern Sudan. J Family Community Med 14: 25–28. [PMC free article] [PubMed] [Google Scholar]
- 12.Sharma VR, 2008. When to seek health care: a duration analysis for malaria patients in Nepal. Soc Sci Med 66: 2486–2494. [DOI] [PubMed] [Google Scholar]
- 13.Department of Child and Adolescent Health and Development, World Health Organization, UNICEF , 2005. Handbook IMCI: Integrated Management of Childhood Illness. Geneva, Switzerland: World Health Organization. [Google Scholar]
- 14.D’acremont V, Kilowoko M, Kyungu E, Philipina S, Sangu W, Kahama-Maro J, Lengeler C, Cherpillod P, Kaiser L, Genton B, 2014. Beyond malaria—causes of fever in outpatient Tanzanian children. N Engl J Med 370: 809–817. [DOI] [PubMed] [Google Scholar]
- 15.Deressa W, Fantahun M, Ali A, 2007. Malaria-related mortality based on verbal autopsy in an area of low endemicity in a predominantly rural population in Ethiopia. Malar J 6: 128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kamat VR, 2006. “I thought it was only ordinary fever!” cultural knowledge and the micropolitics of therapy seeking for childhood febrile illness in Tanzania. Soc Sci Med 62: 2945–2959. [DOI] [PubMed] [Google Scholar]
- 17.Malawi Ministry of Health , 2011. Malaria Strategic Plan 2011–2015 Towards Universal Access. Lilongwe, Malawi: National Malaria Control Programme Community Health Sciences Unit. Available at: https://www.medbox.org/malaria-strategic-plan-2011-2015/download.pdf. Accessed January 29, 2019.
- 18.National Statistical Office (NSO)/Malawi, ORC Macro , 2005. Malawi Demographic and Health Survey 2004. Calverton, MD: NSO/Malawi and ORC Macro. [Google Scholar]
- 19.National Statistical Office (NSO)/Malawi, ICF Macro , 2011. Malawi Demographic and Health Survey 2010. Zomba, Malawi: NSO/Malawi and ICF Macro. [Google Scholar]
- 20.National Malaria Control Programme (NMCP)/Malawi, ICF International , 2012. Malawi Malaria Indicator Survey 2012. Lilongwe, Malawi: NMCP/Malawi and ICF International. [Google Scholar]
- 21.National Malaria Control Programme (NMCP)/Malawi, ICF International , 2015. Malawi Malaria Indicator Survey 2014. Rockville, MD: NMCP/Malawi and ICF International. [Google Scholar]
- 22.Hamooya BM, Chongwe G, Dambe R, Halwiindi H, 2016. Treatment-seeking behaviour for childhood fever among caretakers of Chivuna and Magoye rural communities of Mazabuka district, Zambia: a longitudinal study. BMC Public Health 16: 762. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.National Statistical Office (NSO) [Malawi] and ICF , 2017. Malawi Demographic and Health Survey 2015–16: Zomba, Malawi and Rockville, MD: NSO and ICF. [Google Scholar]
- 24.Malawi Malaria Impact Evaluation Group , 2016. Evaluation of the Impact of Malaria Control Interventions on All-Cause Mortality in Children Under-Five in Malawi Washington, DC: President’s Malaria Initiative. [Google Scholar]
- 25.Awantang G, Babalola S, Koenker H, Fox K, Toso M, Lewicky N, Somah D, Koko V, 2018. Correlates of social behavior change communication on care-seeking behaviors for children with fever: an analysis of malaria household survey data from Liberia. Malar J 17: 105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Adinan J, Damian DJ, Mosha NR, Mboya IB, Mamseri R, Msuya SE, 2017. Individual and contextual factors associated with appropriate healthcare seeking behavior among febrile children in Tanzania. PLoS One 12: e0175446. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Rutstein SO, Rojas G, 2006. Guide to DHS Statistics. Rockville, MD: ICF. [Google Scholar]
- 28.Kassile T, Lokina R, Mujinja P, Mmbando BP, 2014. Determinants of delay in care seeking among children under five with fever in Dodoma region, central Tanzania: a cross-sectional study. Malar J 13: 348. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Espino F, Manderson L, 2000. Treatment seeking for malaria in Morong, Bataan, the Philippines. Soc Sci Med 50: 1309–1316. [DOI] [PubMed] [Google Scholar]
- 30.Wiseman V, Scott A, Conteh L, McElroy B, Stevens W, 2008. Determinants of provider choice for malaria treatment: experiences from the Gambia. Soc Sci Med 67: 487–496. [DOI] [PubMed] [Google Scholar]
- 31.Ntenda PAM, Chuang KY, Tiruneh FN, Chuang YC, 2017. Analysis of the effects of individual and community level factors on childhood immunization in Malawi. Vaccine 35: 1907–1917. [DOI] [PubMed] [Google Scholar]
- 32.Nkoka O, Chuang T-W, Chuang K-Y, Chen Y-H, 2018. Factors associated with insecticide-treated net usage among women of childbearing age in Malawi: a multilevel analysis. Malar J 17: 372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Babalola S, Ricotta E, Awantang G, Lewicky N, Koenker H, Toso M, 2016. Correlates of intra-household ITN use in Liberia: a multilevel analysis of household survey data. PLoS One 11: e0158331. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.WHO , 2016. Recommendations on Antenatal Care for a Positive Pregnancy Experience. Geneva, Switzerland: World Health Organization; Available at: https://www.who.int/reproductivehealth/publications/maternal_perinatal_health/anc-positive-pregnancy-experience/en/. Accessed January 29, 2019. [PubMed] [Google Scholar]
- 35.Nkoka O, Chuang T-W, Chen Y-H, 2018. Association between timing and number of antenatal care visits on uptake of intermittent preventive treatment for malaria during pregnancy among Malawian women. Malar J 17: 211. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Robson K, Pevalin D, 2016. Multilevel Modeling in Plain Language. London, United Kingdom: Sage Publications. [Google Scholar]
- 37.Gupta M, et al. 2003. International differences in in-hospital revascularization and outcomes following acute myocardial infarction: a multilevel analysis of patients in ASSENT-2. Eur Heart J 24: 1640–1650. [DOI] [PubMed] [Google Scholar]
- 38.Ntenda PAM, Chuang Y-C, 2018. Analysis of individual-level and community-level effects on childhood undernutrition in Malawi. Pediatr Neonatol 59: 380–389. [DOI] [PubMed] [Google Scholar]
- 39.Mayala B, Fish TD, Eitelberg D, Dontamsetti T, 2018. The DHS Program Geospatial Covariate Datasets Manual, 2nd edition Rockville, MD: ICF. [Google Scholar]
- 40.Altman DG, Bland JM, 1994. Statistics notes: quartiles, quintiles, centiles, and other quantiles. BMJ 309: 996. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.O’brien RM, 2007. A caution regarding rules of thumb for variance inflation factors. Qual Quant 41: 673–690. [Google Scholar]
- 42.Evans EA, Upchurch DM, Simpson T, Hamilton AB, Hoggatt KJ, 2017. Differences by veteran/civilian status and gender in associations between childhood adversity and alcohol and drug use disorders. Soc Psychiatry Psychiatr Epidemiol 53: 421–435. [DOI] [PubMed] [Google Scholar]
- 43.Masangwi SJ, Ferguson N, Grimason A, Morse T, Kazembe L, 2010. Care-seeking behaviour and implications for malaria control in southern Malawi. South Afr J Epidemiol Infect 25: 22–26. [Google Scholar]
- 44.Kazembe LN, Appleton CC, Kleinschmidt I, 2007. Choice of treatment for fever at household level in Malawi: examining spatial patterns. Malar J 6: 40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Chinkhumba J, Skarbinski J, Chilima B, Campbell C, Ewing V, San Joaquin M, Sande J, Ali D, Mathanga D, 2010. Comparative field performance and adherence to test results of four malaria rapid diagnostic tests among febrile patients more than five years of age in Blantyre, Malawi. Malar J 9: 209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Mathanga DP, Walker ED, Wilson ML, Ali D, Taylor TE, Laufer MK, 2012. Malaria control in Malawi: current status and directions for the future. Acta Trop 121: 212–217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Ansari S, Soltero EG, Lorenzo E, Lee RE, 2017. The impact of religiosity on dietary habits and physical activity in minority women participating in the Health is Power (HIP) study. Prev Med Rep 5: 210–213. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Chege PM, Kimiywe JO, Ndungu ZW, 2015. Influence of culture on dietary practices of children under five years among Maasai pastoralists in Kajiado, Kenya. Int J Behav Nutr Phys Act 12: 131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Tan KL, 2011. Factors associated with exclusive breastfeeding among infants under six months of age in peninsular Malaysia. Int Breastfeed J 6: 2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Kretchy I, Owusu-Daaku F, Danquah S, 2013. Spiritual and religious beliefs: do they matter in the medication adherence behaviour of hypertensive patients? Biopsychosoc Med 7: 15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Tegegne M, Sileshi S, Benti T, Teshome M, Woldie H, 2017. Factors associated with minimal meal frequency and dietary diversity practices among infants and young children in the predominantly agrarian society of Bale zone, southeast Ethiopia: a community based cross sectional study. Arch Public Health 75: 53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Pillai RK, Williams SV, Glick HA, Polsky D, Berlin JA, Lowe RA, 2003. Factors affecting decisions to seek treatment for sick children in Kerala, India. Soc Sci Med 57: 783–790. [DOI] [PubMed] [Google Scholar]
- 53.WHO , 2013. Pocket Book of Hospital Care for Children: Guidelines for the Management of Common Childhood Illnesses. Geneva, Switzerland: World Health Organization. [PubMed] [Google Scholar]
- 54.Essomba NE, Kedy Koum DC, Adiogo D, Ngwe MI, Coppieters Y, 2015. Use of oral rehydration therapy in the treatment of childhood diarrhoea in Douala, Cameroon. Malawi Med J 27: 60–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Masangwi S, Ferguson N, Grimason A, Morse T, Kazembe L, 2016. Care-seeking for diarrhoea in southern Malawi: attitudes, practices and implications for diarrhoea control. Int J Environ Res Public Health 13: 1140. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Ibrahim IR, Palaian S, Ibrahim MI, 2018. Assessment of diarrhea treatment and counseling in community pharmacies in Baghdad, Iraq: a simulated patient study. Pharm Pract 16: 1313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Pemberton S, Humphris P, 2016. Locality, Neighbourhood and Health: A Literature Review, IRiS Working Paper Series, No. 13/2016 (UPWEB Working Paper Series, No. 2/2016) Birmingham, England: Institute for Research into Superdiversity. [Google Scholar]
- 58.Evans CR, Williams DR, Onnela JP, Subramanian SV, 2018. A multilevel approach to modeling health inequalities at the intersection of multiple social identities. Soc Sci Med 203: 64–73. [DOI] [PubMed] [Google Scholar]
- 59.Menon MP, Njau JD, McFarland DA; Uganda Malaria Indicator Survey Technical Working Group , 2016. Cost and predictors of care-seeking behaviors among caregivers of febrile children-Uganda, 2009. Am J Trop Med Hyg 94: 932–937. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Kahabuka C, Kvale G, Hinderaker SG, 2013. Care-seeking and management of common childhood illnesses in Tanzania—results from the 2010 demographic and health survey. PLoS One 8: e58789. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61.Hwang J, Graves PM, Jima D, Reithinger R, Kachur SP; the Ethiopia MISWG , 2010. Knowledge of malaria and its association with malaria-related behaviors—results from the malaria indicator survey, Ethiopia, 2007. PLoS One 5: e11692. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62.Akbarinejad F, Soleymani M, Shahrzadi L, 2017. The relationship between media literacy and health literacy among pregnant women in health centers of Isfahan. J Educ Health Promot 6: 17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63.Zimmerman E, Woolf SH, 2014. Understanding the Relationship between Education and Health. Discussion Paper. Washington, DC: Institute of Medicine; Available at: https://nam.edu/wp-content/uploads/2015/06/BPH-UnderstandingTheRelationship1.pdf. Accessed January 29, 2019. [Google Scholar]
- 64.Sabates R, Feinstein L, 2006. The role of education in the uptake of preventative health care: the case of cervical screening in Britain. Soc Sci Med 62: 2998–3010. [DOI] [PubMed] [Google Scholar]
- 65.Baumann E, Czerwinski F, Reifegerste D, 2017. Gender-specific determinants and patterns of online health information seeking: results from a representative German health survey. J Med Internet Res 19: e92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66.Sridharan S, Dey A, Seth A, Chandurkar D, Singh K, Hay K, Gibson R, 2017. Towards an understanding of the multilevel factors associated with maternal health care utilization in Uttar Pradesh, India. Glob Health Action 10: 1287493. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 67.Chibwana AI, Mathanga DP, Chinkhumba J, Campbell CH, 2009. Socio-cultural predictors of health-seeking behaviour for febrile under-five children in Mwanza-Neno district, Malawi. Malar J 8: 219. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 68.Anselmi L, Lagarde M, Hanson K, 2015. Health service availability and health seeking behaviour in resource poor settings: evidence from Mozambique. Health Econ Rev 5: 26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 69.Kenny A, et al. 2015. Remoteness and maternal and child health service utilization in rural Liberia: a population-based survey. J Glob Health 5: 020401. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 70.Sonkong K, Chaiklieng S, Neave P, Suggaravetsiri P, 2015. Factors affecting delay in seeking treatment among malaria patients along Thailand-Myanmar border in Tak Province, Thailand. Malar J 14: 3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 71.Robert SA, 1998. Community-level socioeconomic status effects on adult health. J Health Soc Behav 39: 18–37. [PubMed] [Google Scholar]