Longitudinal Depressive and Anxiety Symptoms of Adult Injury Patients in Kenya and their Risk Factors

Abstract

Background:

Injuries account for a significant proportion of the health and economic burden for populations in low- and middle-income countries. However, little is known about psychological distress trajectories amongst injury survivors in low- and middle-income countries.

Methods:

Adult injury patients (n=644) admitted to Kenyatta National Hospital in Nairobi, Kenya, were enrolled and interviewed in the hospital, and at 1, 2–3, and 4–7 months after hospital discharge through phone to assess depressive and anxiety symptoms and level of disability. Growth mixture modeling was applied to identify latent trajectories of depressive and anxiety symptoms.

Results:

Elevated depressive and moderate-level anxiety symptoms (13%) and low depressive and anxiety symptoms (87%) trajectories were found between hospitalization and up to seven months after hospital discharge. Being female, prior trauma experience, longer hospitalization, worse self-rated health status while in the hospital, and lack of monetary assistance during hospitalization were associated with the elevated symptoms trajectory. The higher symptoms trajectory associated with higher disability levels after hospital discharge and significantly lower proportion of resuming daily activities and work.

Conclusion:

The persistence of elevated depressive symptoms and associated reduced functioning several months after physical injury underscores the importance of identifying populations at risk for preventive and early interventions.

Introduction

Injury is a major public health concern and accounts for one-tenth of the disability-adjusted life years lost globally, a measure that combines the burden of mortality and morbidity (1). Major causes of injuries include road traffic collisions, self-harm, falls, and interpersonal violence (2). Despite a declining global trend in the rates of disability-adjusted life years lost due to injury compared to 1990, the injury rates among children and adults before age 50 in low- and middle-income regions remain much higher than in high-income regions (2,3). In particular, the high rates of injury among younger adults in Sub-Saharan Africa largely comprised of road traffic injuries and interpersonal violence (2). Studies have suggested the increase in motorization outpaced infrastructure development and law enforcement and contributed to the greater burden of road traffic injuries in low- and middle-income countries; while poverty, social isolation, and access to firearms were some underlying causes of violence (3).

In addition to the prevention of injury incidence and mortality, there is a growing need to understand the consequences of injury among survivors in low- and middle-income countries. Injury not only impairs physical health but also has psychological and social consequences. Emerging research has found various mental health consequences of injury, including post-traumatic stress disorder, depression, and anxiety (4). Studies on the onset of depression and anxiety after injury have been conducted predominantly in high-income contexts (4), which found prevalent and elevated depressive symptoms among injury survivors several months to years after injury (5–7). However, survivors may have different trajectories of psychological distress in response to the incident which may not be captured in binary categories (8). Previous studies of individual differences of depressive symptoms after injury in high-income settings included survivors of specific injuries, including spinal cord injury (9,10), traumatic brain injury (11–13), burns (14), hip fracture (15), as well as general traumatic injury (16). These studies identified between three to five distinct trajectories, some of which had a stable course (low, medium, and persistent high), while other courses of changing severity (recovery, delayed onset) were found in some studies (12,16). However, it is unclear whether these trajectories and their associated factors may be different in the lower resource context.

Disability is a concept that includes dimensions of functionality, activity, and contextual factors. While the severity and specific nature of the physical injury have traditionally been the focus regarding disability, studies have shown both physical and psychological consequences of injury also contribute to disability (6,17). In fact, a study on trauma patients in Australia found that psychological symptoms accounted for the most variation in disability one year after injury (17). While these findings underlined the association between psychological symptoms and disability, understanding distinct trajectories of psychological symptoms and associated disability would enable identifying subgroups within the injury population that may be at increased risk, as well as specific characteristics and malleable mechanisms for preventive measures. Despite a higher burden of injuries in low- and middle-income countries (2,3), little is known about trajectories of depressive and anxiety symptoms and their associated level of disability in the low-income contexts to inform and target prevention efforts.

Kenya is a lower middle-income country that is rapidly developing, with a predominantly young population and growing number of overall physical injuries, especially road traffic injuries and interpersonal violence (18). A previous study in Kenya found over 30% of patients who had general surgery or orthopedic injury had general depression and over 10% had general anxiety (19). More general types of injury were considered in a recent cross-sectional study among university students in Nairobi which found elevated depressive symptoms correlated with serious injuries in the past 12 months (20). However, these studies were cross-sectional and cannot examine how depressive and anxiety symptoms develop over time and the persistency of these symptoms.

This study aimed to estimate the trajectories of depression and anxiety symptoms between hospitalization and up to 7 months after hospital discharge among adult injury survivors in a predominantly urban population in Kenya. We hypothesized that the majority of the respondents would develop minimal depressive and anxiety symptoms, with potential differing levels of elevated depressive and anxiety symptoms which would associate with a higher level of disability. Our second aim was to examine risk and protective factors that might explain trajectory membership. We included predictors measured at baseline that were identified in the literature to influence distress symptoms, including demographic characteristics (sex, education, marital status, rural residence), injury characteristics (injury severity, cause of injury, intent of injury), and social (financial stability, financial assistance) and psychosocial (exposure to potentially traumatic events) factors (21). We hypothesized that these risk factors would associate with elevated levels of depressive and anxiety symptoms. Our third aim was to estimate the associated disability and functioning at among individuals with elevated depression and anxiety symptoms, and hypothesized that higher disability and worse functioning would associate with individuals with elevated depression and anxiety symptoms.

Methods

Study Population, Recruitment and Enrollment

The current study included adults (age 18 years or older) recruited to participate in a longitudinal study of Health, Economic and Long-term Social Impact of Injuries in Kenya. Participants were recruited between May and December 2015 in Nairobi, Kenya at Kenyatta National Hospital. This hospital was selected as the recruitment site because it is one of the two public hospitals that provide the highest level of care in Kenya and serves as the largest referral and teaching hospital in the country. Recruitment was conducted in the orthopedics wards, general surgical wards, ear and throat wards, head and abdominal wards, burns ward and burns unit. Patients were recruited to the study based on the following criteria: 1) they had experienced one or more unintentional injuries or assault; 2) required hospital admission of at least 24 hours; 3) were 18 years or older; 4) were able to communicate in Swahili or English; 5) resided in Kenya; 6) intended to stay in Kenya after hospital discharge; 7) were able to provide informed consent and to be interviewed at the time of enrollment, based on consciousness and coherence in the cases of head injury. For the current study, patients were excluded if they were admitted to the hospital due to self-harm, if they self-reported having been diagnosed with any mental disorder(s) pre-injury, or if they did not survive through hospitalization.

Eligible patients were consecutively enrolled during weekdays over a seven-month period. Of 1038 patients approached for recruitment to the study, 1004 (96.7%) provided informed consent to participate in the parent study (figure 1). Of those, 961 met the inclusion criteria for the current study. Due to issues with data quality of one of the three data collectors, participants recruited and followed up by this data collector were excluded from this study (n=317). Those excluded were more likely to be female, residing in a rural area, and injured from an unintentional cause (p<0.05). This resulted in a final sample of 644 patients. First follow-up at 1-month after hospital discharge included 338 participants (35.2% of those who met inclusion criteria), second follow-up at 2–3 months post-hospital discharge included 378 (39.3% of those who met inclusion criteria) participants, and third follow-up at 4–7 months after hospital discharge included 394 (41.0% of those who met inclusion criteria) participants. In total, of the 644 participants included in baseline, 167 (25.9%) were lost-to-follow-up after baseline interviews, and 397 (61.6%) completed at least 2 follow-up interviews post-hospital discharge. Participants who were lost-to-follow-up after baseline interviews did not statistically significantly differ from those who completed at least 1 follow-up by age group, education level, type of occupation, marital status, residence, estimated injury severity level, household savings, having borrowed money for treatment, prior exposure to trauma, duration of hospital stay, having household savings or assets, in-hospital anxiety and depressive symptoms. Those who were lost-to-follow-up after baseline assessments were more likely to be: male (27.9% of males vs. 14.1% of females, p=0.005), injured by assault (36.2% of assault patients vs. 23.1% of unintentional injury patients, p=0.002), and have no medical insurance coverage (29.5% of without medical insurance patients vs. 14.7% of medical insurance covered patients, p<0.001).

Figure caption 1.

Flowchart of injury patient recruitment and attrition, recruited between May and December 2015 in Nairobi, Kenya.

Trained data collectors obtained informed consent and conducted baseline assessments in the hospital. Baseline assessments included an initial interview upon enrollment to obtain demographic information, information about the injury event, disability level pre-injury, prior exposure to trauma, and a self-report measure on depression and anxiety symptoms since the injury event. An exit interview was conducted with each participant regarding in-hospital payment, expenses and assistance received. Participants were asked to provide their cellphone number and number of their next of kin or neighbor. Subsequent follow-up interviews post-hospital discharge were conducted by phone and included questions on depressive and anxiety symptoms and disability. For participants who could not be reached at first, data collectors tried contacting them at least three other times at different times of the day and on both weekdays and weekends, followed by contacting their next of kin or neighbor. Participants who could not be reached at the first follow-up were contacted again in subsequent follow-ups.

This study received Institutional Review Board approval from Johns Hopkins University Bloomberg School of Public Health in the U.S. and the Ethics Review Committee of University of Nairobi and Kenyatta National Hospital in Kenya.

Measures

Depression and anxiety symptoms.

We used the Hopkins Symptoms Checklist to measure anxiety and depression symptoms. Hopkins Symptoms Checklist has previously demonstrated good interrater reliability, test-retest reliability, internal consistency and content validity in Tanzania (22), and has been used frequently in the eastern Sub-Saharan Africa region (23,24). The depression subscale performance was robust across eight low- and middle income countries, including Uganda and Rwanda (25). Severity was ranked on a Likert scale from 1 (not at all) to 4 (extremely) with recall of the past week’s symptoms. As commonly reported in the literature (26), mean scores were generated using the anxiety subscale (10 items) and depressive subscale (15 items). Cronbach’s alpha of the depression and anxiety subscale in the current sample measured in hospital was 0.71 and 0.65 respectively.

Potentially Traumatic Events.

The 10-item self-report Brief Trauma Questionnaire was used to assess the exposure to various types of potentially traumatic events. Derived from the Brief Trauma Interview (27), the Brief Trauma Questionnaire provides a complete assessment of criterion A of post-traumatic stress disorder in DSM-5 and has been used among nurses and civilians to assess trauma exposure (28). The instrument includes common types of potentially traumatic events but remains succinct and feasible for implementation. Questions on whether the participant was seriously injured and whether the participant thought his or her life was in danger or might be seriously injured were included for each type of trauma. Exposure to any potentially traumatic events was coded as binary indicator (no exposure=0; any exposure=1).

Level of Disability, resumption of work, and normal daily activities.

The World Health Organization Disability Assessment Schedule 2.0 12-item self-report instrument was used to assess participant’s experiences of functional impairment or disability associated with a health condition. The instrument has been used to measure disability among general populations in different low- and middle-income countries (29), as well as those with medical conditions (30). Its psychometric properties have been evaluated across cultures which supported the unidimensional property of the scale (31). Items were ranked based on the level of difficulty due to health conditions and ranged between 0 (none) and 4 (extreme or cannot do). Scoring of the scale was computed according to the 12-item instrument scoring sheet to convert the item scores to global disability scores that ranged between 0 (no disability) and 100 (full disability). Cronbach’s alpha in the current sample was 0.82. Resumption of work and normal daily activities were each assessed by a single question at each follow-up interview. Response to return to work and resuming normal daily activities were binary.

Injury severity and injury-related covariates.

Information of injury-related covariates and hospitalization were obtained from the patient’s medical record. Diagnosis of injury, including anatomy and pathology of the injury, was recorded from each participant’s medical record by a medical resident. Up to three injuries per patient were recorded on the baseline questionnaire. An estimated abbreviated injury scale was generated based on the anatomy and pathology of the injury. An estimated Injury Severity Score was then calculated from the estimated abbreviated injury scale for each patient (32). The estimated Injury Severity Score has been the most commonly used measure in quantifying severity of an injury among trauma registries in LMICs, including Kenya, with a satisfactory discriminating ability on in-hospital deaths (Area Under Curve: 0.76) (33). Injury severity was categorized as slight severity (estimated Injury Severity Score <9), moderate severity (estimated Injury Severity Score 9–15), and severe (estimated Injury Severity Score >15) (34). Anatomy and pathology of injury, as recorded in the participant’s medical record, was used to determine the presence of mild traumatic brain injury. Participants were also asked about the injury event, including the intent and cause of the injury, and their self-rated health state during hospitalization during the baseline interview. Finally, at the hospital exit interview, participants were asked if they received any type of assistance from government or non-governmental organizations and about the type(s) of assistance they received. They were also asked if they had health insurance coverage and if they borrowed money or sold any assets to pay for the hospital fees.

Sociodemographic variables.

At the baseline interview, participants were asked their age, marital status, highest level of education, type of occupation, and residence (rural/urban).

Translation of measures.

We used the translated Swahili version of Hopkins Symptoms Checklist adapted for Tanzania in this study. Other measures, including the Brief Trauma Questionnaire and World Health Organization Disability Assessment Schedule 2.0, were translated to Swahili by a certified translator in Kenya. The translated measures were then reviewed and modified by the local supervisor and data collectors during training for data collection, after group discussion, and upon piloting of the questionnaire. All questions were administered by the interviewer.

Statistical Analyses

We used piecewise growth mixture modeling to examine trajectories of depression and anxiety symptoms after injury. Parameters were estimated using a full-information maximum likelihood estimator with robust standard errors and fixed within-class variance for growth factors. Unconditional trajectory models were first estimated with progressing numbers of classes, and the best fitting unconditional trajectory model was identified based on model convergence, the Bayesien Information Criterion, the sample size adjusted Bayesien Information Criterion, the Vuong-Lo-Mendell-Rubin likelihood ratio test, the estimated class size for stable parameter estimates, as well as interpretability of the solutions (35).

Covariates were then integrated into the best fitting unconditional model as auxiliary variables to independently assess the relationship between the potential predictor auxiliary variables and the class membership (36). The relationship between the latent classes and the following covariates were examined: age at admission, sex, highest education level attained, marital status, residence (rural or urban), intent of injury (assault vs. unintentional), mechanism of injury (road traffic injury, burns, vs. other), injury severity level, medical insurance coverage, financial status (household savings or assets vs. none, borrowed money for hospital treatment vs. not), assistance received from government or non-governmental organizations, duration of hospitalization, and prior trauma exposure measured by the Brief Trauma Questionnaire. Variables that were statistically significant related to class membership, and covariates that have been previously found to be associated with depression and anxiety were retained for inclusion in the multivariate mixture model. Models were controlled for the data collector to account for potential similarities within the same interviewer. The manual 3-step approach of auxiliary variables was used to estimate the multivariate mixture model (36). Model fit indexes, including Bayesien Information Criterion and the sample size adjusted Bayesien Information Criterion, were compared among models with various covariates to identify the final conditional model. Missing covariates were list-wise deleted (n=51). To assess the association of latent class membership with a global measure of disability, resumption of normal daily activities, and resumption of work, these variables were assessed as distal auxiliary variables using automatic BCH methods (37). All analyses were conducted using Mplus version 7.2 (38).

Results

Participant characteristics

Table 1 shows the demographic and injury characteristics of the study sample. Participants were predominantly male (85.6%), and the median age of participants was 31 years (IQR: 26 – 40). Participants were mostly working as private employees (40.4%) or self-employed (28.7%). The majority of the patients lived in urban areas (93.8%). A higher proportion of men having at least secondary / high school education, work as a private employee, and a marital status of married.

Table 1.

Demographic and injury characteristics of injury patients enrolled in the study, Kenya (n=644)

	Female (n=92)		Male (n=551)		Total (n=644)
	N	%	N	%	N	%
Age (chi2(3) = 3.2178, p = 0.359)
18–29 years	35	38.04	246	44.65	281	43.7
30–44 years	40	43.48	227	41.20	267	41.52
45–59 years	14	15.22	71	12.89	85	13.22
60 years and above	3	3.26	7	1.27	10	1.56
Highest education level (chi2(7) = 21.5116, p= 0.003)
None	2	2.17	11	1.99	13	2.02
Some primary school	19	20.65	53	9.60	72	11.18
Completed primary school	30	32.61	156	28.26	186	28.88
Secondary/high school	22	23.91	256	46.38	278	43.17
Technical/vocational school	6	6.52	27	4.89	33	5.12
Some college, college, or higher levels	13	14.13	49	8.88	62	9.63
Marital Status (chi2(4) = 42.5793, p = 0.000)
Single	32	34.78	183	33.21	215	33.44
Married	40	43.48	337	61.16	377	58.63
Widowed	9	9.78	7	1.27	16	2.49
Divorced/separated	11	11.96	24	4.35	35	5.44
Residence: urban / rural (chi2(1) = 0.0178, p = 0.894)
Rural	6	6.52	34	6.16	40	6.21
Urban	86	93.48	518	93.84	604	93.79
Intent of injury (chi2(1) = 1.0391, p = 0.308)
Unintentional	76	82.61	430	77.9	506	78.57
Assault	16	17.39	122	22.1	138	21.43
Cause of unintentional injury (chi2(9) = 22.8550, p = 0.007)
Road traffic	41	53.95	269	62.56	310	61.26
Fall	16	21.05	70	16.28	86	17.00
Burn	13	17.11	23	5.35	36	7.11
Sharp object	4	5.26	23	5.35	27	5.34
Explosive	0	0.00	3	0.70	3	0.59
Blunt object	0	0.00	22	5.12	22	4.35
Electrocution	0	0.00	11	2.56	11	2.17
Others	2	2.63	9	2.09	11	2.17
Estimated injury severity level (chi2(2) = 4.3858, p = 0.112)
eISS <9	44	51.76	213	40.73	257	42.27
eISS 9–15	38	44.71	272	52.01	310	50.99
eISS >15	3	3.53	38	7.27	41	6.74
Duration of hospital stay (chi2(4) = 9.7088, p = 0.046)
Less than 2 weeks	33	35.87	214	38.91	247	38.47
2–4 weeks	12	13.04	104	18.91	116	18.07
4–6 weeks	22	23.91	76	13.82	98	15.26
6–8 weeks	5	5.43	58	10.55	63	9.81
More than 8 weeks	20	21.74	98	17.82	118	18.38

More than one-fifth of the patients were injured as a result of intentional physical assault (21.4%). Among unintentional injuries, road traffic injuries were the most common (61.3%, male: 62.6%, female: 54.0%), followed by falls (male: 16.3%, females: 21.1%). About a tenth of all participants (n= 67) had mild traumatic brain injury. More than one-third of patients were hospitalized for less than 2 weeks (38.5%), while nearly one-fifth of patients remained hospitalized for more than 2 months (18.4%).

Latent depression and anxiety symptoms trajectories

Table 2 shows the model fit statistics of piecewise latent growth curve models. A two-class enumeration was selected based on Bayesien Information Criterion and the sample size adjusted Bayesien Information Criterion with the largest decrease in a two-class solution. The two-class solution was also statistically distinct from the one-class solution, as indicated by the Vuong-Lo-Mendell-Rubin likelihood ratio test.

Table 2.

Model statistics of unconditioned latent class enumeration of depressive and anxiety symptoms (HSCL-25) of injury patients between baseline and 4–7 months post-discharge (n=644)

No. of class(es)	df	LL	AIC	BIC	Sample-size adjusted BIC	VLMR LRT		LMR-LRT		Entropy	Smallest class n (%)
						2*LL	p-value	2*LL	p-value
1	13	141.79	−257.58	−199.50	−240.78		NA		NA	NA	NA
2	19	482.18	−926.35	−841.47	−901.79	680.77	0.017	663.67	0.019	0.847	82 (12.7%)
3	25	587.48	−1124.96	−1013.27	−1092.64	210.61	0.276	205.32	0.282	0.843	12 (1.9%)
4	31	681.41	−1300.82	−1162.32	−1260.74	187.86	0.200	183.14	0.206	0.842	6 (0.9%)

Note: df=degrees of freedom; LL=log-likelihood; AIC=Akaike information criterion; BIC=Bayesian information criterion; a-BIC=adjusted BIC; VLMR LRT=Vuong-Lo-Mendell-Rubin likelihood ratio test; LMR LRT=Lo-Mendell-Rubin likelihood ratio test.

Figure 2 displays the depression and anxiety trajectories with a 2-class model: elevated depressive symptoms with moderate anxiety symptoms (12.7%) and low depressive and anxiety symptoms (87.3%). Individuals who were likely to be in the elevated symptoms class had moderate depressive symptoms at baseline (mean: 1.49, p<0.001). In the elevated symptoms class, depressive symptoms increased by 1-month post hospital discharge (slope: 0.22, p<0.001) then remained stable (slope: 0.015, p=0.868) between 1-month and 4–7 months post discharge, while anxiety symptoms remained stable (slope: −0.026, p=0.346). The majority of participants (87.3%) had low levels of depressive and anxiety symptoms while in the hospital (Hopkins Symptoms Checklist depressive subscale: 1.24, Hopkins Symptoms Checklist anxiety subscale: 1.13), which remained stable after discharge (p>0.1).

Figure caption 2.

Trajectories of depression and anxiety symptoms of injury patients in Kenya between baseline in hospital and 4–7 months after hospital discharge (2-class model, n=644).

Latent regression with elevated depression and moderate anxiety symptoms class

Table 3 shows the results of a multivariate logistic regression on the mixture model. Similar to the bivariate analysis (available on request), being female remained strongly associated with the elevated depression and moderate anxiety symptoms class (aOR: 5.74, 95% CI: 2.23–14.35). Prior exposure to trauma (aOR: 3.53, 95% CI: 1.47–8.26), and hospitalization duration (aOR: 1.49, 95% CI: 1.14–1.94) were also associated with being in the elevated depression and moderate anxiety symptoms class. Having received monetary assistance during hospitalization (aOR: 0.27, 95% CI: 0.07–0.94) and better self-rated health state during hospitalization (aOR: 0.77, 95% CI: 0.60–0.98) were associated with lower odds of being in the elevated depression and moderate anxiety symptoms class. The associations with estimated injury severity level (aOR: 1.09, 95% CI: 0.60–1.94) was not statistically significant in the multivariate mixture model.

Table 3.

Multivariate logistic regression to estimate the associations between baseline characteristics and elevated depression and anxiety symptoms trajectories of injury patients between baseline and 4–7 months post-discharge (n=644)

Multivariate association with elevated depression and moderate anxiety symptoms class
Covariates	aOR	p-value	95% CI
Female (vs. male)	5.74	0.000	2.23	14.35
Age (increment of 10 years)	0.75	0.127	0.52	1.07
Residence (Rural vs. urban)	2.19	0.227	0.61	7.55
Self-rated health status during hospitalization	0.77	0.038	0.60	0.98
Mild traumatic brain injury (vs. other injury)	1.38	0.613	0.39	4.72
Estimated injury severity level (severe, moderate, mild)	1.09	0.780	0.60	1.94
Intent of injury (assault vs. unintentional)	0.88	0.794	0.35	2.19
Prior exposure to trauma	3.53	0.005	1.47	8.26
Duration of hospitalization (increment of half a month)	1.49	0.004	1.14	1.94
Received monetary assistance during hospitalization	0.27	0.046	0.07	0.94
Proportion of medical bill covered by insurance (increment of 10%)	0.94	0.241	0.85	1.04

Note: aOR=adjusted odds ratio.

Association between symptoms trajectories and disability, normal activities and work

Results in table 4 indicate individuals with elevated depression and moderate anxiety symptoms had almost three times higher disability levels between 1 and 4–7 months after hospital discharge (4–7 months mean of World Health Organization Disability Assessment Schedule 2.0: 28.4 among elevated depressive and moderate anxiety symptoms trajectory class, vs. 10.3 among low depressive and anxiety symptoms trajectory class, p<0.001). While nearly half the individuals with low depression and anxiety symptoms resumed some normal daily activities by 1 month after hospital discharge (43.5%), only 7.9% of individuals with elevated depression and moderate anxiety symptoms did so. By 4–7 months after hospital discharge, about a third of individuals with elevated depression and moderate anxiety symptoms resumed some normal daily activities, but only 7.5% had resumed work compared to 44.3% of individuals with low depression and anxiety symptoms.

Table 4.

Level of disability, resumption of normal daily activities, and resumption of work comparing between elevated depression and anxiety symptoms class and low symptoms class among injury patients at 1, 2–3, and 4–7 months after hospital discharge in Kenya (n=644)

		Elevated depression and moderate anxiety symptoms class		Low depression and anxiety symptoms class		Overall test between class
		Mean	SE	Mean	SE	Chi-square	P-value
Disability level	1-month	42.7	2.7	24.1	1.0	38.68	<0.001
	2–3 months	37.3	2.3	15.2	0.9	72.98	<0.001
	4–7 months	28.4	3.3	10.3	0.7	27.67	<0.001
		Proportion	SE	Proportion	SE	Chi-square	P-value
Resumed normal daily activities	1-month	0.079	0.06	0.435	0.03	108.49	<0.001
	2–3 months	0.188	0.06	0.634	0.03	50.81	<0.001
	4–7 months	0.356	0.78	0.778	0.02	262.35	<0.001
		Proportion	SE	Proportion	SE	Chi-square	P-value
Resumed work	1-month	0.044	0.04	0.107	0.02	14.20	<0.001
	2–3 months	0.031	0.03	0.296	0.03	197.10	<0.001
	4–7 months	0.075	0.05	0.443	0.03	28.24	<0.001

Discussion

This study provides evidence for the burden of post-injury depressive and anxiety symptoms several months beyond hospital discharge in a predominantly urban male population in Kenya. As hypothesized, we found the majority of injury survivors had low depressive and anxiety symptoms after injury, which was consistent with previous research (16,39). Our results identified about 13% of the sample with elevated depressive and moderate anxiety symptoms trajectories after hospitalization. Unlike in other studies on depressive symptoms trajectories, we did not observe a distinct pattern of improved symptoms class or a delayed pattern. The lack of improved symptoms class and delayed pattern may be due to the shorter follow-up period compared with other studies on injury survivors (11,16). Being part of this elevated symptoms class was associated with three times higher levels of disability and strongly reduced resumption of normal activities and work. Objective injury factors, including cause and severity of an injury, did not predict the trajectory of post-injury depression and anxiety, similar to prior studies (40,41).

The increase in persistent depressive symptoms among this subgroup of injury survivors one month post-hospital discharge underlines the need for early follow-up screening for depressive symptoms after discharge. Additionally, consistent with the literature in developed settings, being female and having a history of potentially traumatic events exposure are strongly associated with elevated depressive and anxiety symptoms (21,42). Health care providers following up with injury survivors should screen for depressive and anxiety symptoms early in the rehabilitation process, while paying special attention to women and people with a potential traumatic exposure history. Screening tools for psychological distress in traumatic injury populations have been developed considering the risk factors of subsequent development of psychopathology in high-income settings (43). Such tools may be adapted for use in low-resource settings through cultural adaptation, validation, and reliability assessment, and can be informed by risk factors identified in the current study. Effective screening by healthcare providers can assist in identifying people who may benefit from evidence-based selective prevention and treatment interventions, such as a trauma focused cognitive behavioral therapy (44). Studies have demonstrated that identifying injury survivors with higher risk of psychological distress provided an important opportunity to target early psychological intervention to prevent the development of mental disorders in developed settings (45,46). Healthcare managers and providers should consider incorporating evidence-based culturally adapted early psychosocial interventions in rehabilitation and outpatient clinics that provide medical follow-up for injury survivors to prevent associated disability (47).

Our study found duration of hospitalization was associated with being in the elevated depressive and moderate anxiety symptoms trajectory. Compared with other longitudinal studies on post-injury mental health issues carried out in developed countries, injury patients in our study had much lengthier hospitalizations (5,48,49). This difference in hospitalization duration may be partially explained by the time intervals in treatment, as shown in the study comparing time intervals of fractured femurs treatment in low- and middle-income countries and high-income countries (50). The authors found significantly longer treatment in low- and middle-income countries, illustrating the stark difference in resources and treatment availability between high-income and low-income settings (50,51).

Hospitalization duration also has social and economic implications. Prolonged hospitalization does not only prevent patients from returning to normal activities and work but also places a large burden on the patient’s family (as patients are not able to fulfill family roles) and produces an added financial burden. While the measure on hospitalization duration did not distinguish these individual potential aspects, the large reduction in risk of having elevated depressive and moderate anxiety symptoms trajectory with monetary assistance from the government or non-governmental organization during hospitalization underscores the financial stress affecting one’s post-injury mental health. Financial support for injury survivors has been shown to improve return to work times and socio-economic outcomes in high-income settings (52). Such mechanism should be strengthened by the government and non-governmental organizations in this low-resource context to prevent the exacerbation of emotional distress and disability after injury.

The notably large differences in disability between the two symptom level groups on disability is concerning. The elevated depressive and moderate anxiety symptoms group had a statistically significantly higher level of disability between one and several months after hospital discharge. The patterns of the relationship between psychological symptoms and disability are consistent with previous findings that mental health is strongly associated with disability among injured populations (17,53). In addition, this study also found that a strikingly low proportion of participants resumed normal activities and work several months after discharge compared to the low symptoms class, a critical finding in an already financially stressed population. This underlines the importance to prevent, as well as early identification of psychological distress symptoms among injury patients and ensuring access to evidence-based, culturally adapted psychosocial intervention.

We identified several limitations in this prospective study. First, over a quarter of participants were lost-to-follow-up after hospital discharge. A large proportion of those participants were males, injured from assault, without medical insurance and no household savings or assets. These differences in drop-out may have affected the results and the estimation of risk factors. Second, we excluded data from one of the three data collectors due to data quality issues, which limited the generalizability of the data. Third, we used a different modality of assessment between baseline (in-person) and follow-up interviews (by phone). Despite the change in modality, evidence has indicated good agreement between the two modalities in other populations of similar age (54). Our study followed the recommendations of Evans et al. to establish good personal contact between researcher and subjects, including having the same interviewer consistently follow up with the participant to establish rapport. Fourth, as the Injury Severity Score was unavailable in the Kenya setting, an estimated Injury Severity Score was applied using clinical diagnosis information extracted from medical records. The measure may not fully capture the injury severity due to the lack of diagnostic resources in the low-income setting. Additionally, Also, this study included a largely male and urban sample, which may not generalize to females and rural residents. However, it was similar to the general injury population in the hospital from a previous surveillance study and had similar demographic characteristics compared to other public hospitals (18). Future study may include both rural and urban hospitals, investigate retaining assault victims, and consider longer follow up with injury survivors.

Conclusion

We found that 13% of a young, urban, largely male group of injury survivors in Kenya had persisting elevated depression and moderate anxiety symptoms from hospitalization to several months after discharge. This longitudinal elevated psychological distress was associated with significantly high level of disability, lack of resuming normal activities and resuming work several months after hospital discharge. Being females, prior exposure to potentially traumatic events, financial stress, and longer period of hospitalization in this predominantly low-income population were more likely to have elevated psychological distress after injury. While the majority of injury survivors had low level of distress over time, our findings highlight the importance to enable health providers to identify psychological distress early in the physical treatment process and ensure access to evidence-based, culturally adapted psychosocial intervention to reduce disability.