Abstract
Background:
In community settings, lack of availability of care, nonadherence to medication, high cost of treatment, and poor outcomes of treatment are the core problems in noncommunicable disease (NCD) management. Is there an effect of the community-based SPARSH program on treatment adherence, treatment expenditure, and control among selected NCDs? What are the reasons for low treatment adherence to treatment from patients’ point of view?
Methodology:
It was a community-based explanatory type of mixed method evaluation design, where major quantitative (cross-sectional survey) design followed a minor qualitative (free list) component. A representative sample of 850 NCD patients were selected by two-stage cluster sampling. Structured interviews were done to collect information on history of tobacco and alcohol consumption, treatment adherence, measurement of obesity, blood sugar, and blood pressure levels among known NCD patients. A directed acyclic graph was developed, and path analysis was performed using STATA software. Free list was done with patients who were found to have low adherence to treatment. The study was approved by the Institutional Ethics Committee.
Results:
As revealed from path analysis, NCD patients enrolled in the SPARSH program had better treatment adherence, which ultimately led to control of hypertension and diabetes
.
Ninety percent of patients enrolled in SPARSH had high adherence to treatment, and 90% of them had control in their NCD status. Patients with low adherence to medication reported three major reasons for it: 1) forgetting to take the medicines due to a busy schedule, 2) lack of money to purchase medicines, and 3) feeling better, so felt not necessary to take medicines.
Conclusion:
The SPARSH program was found to be effective to control hypertension and diabetes among enrolled patients for its services by improving treatment adherence among them. The key strategy of the program based on home visits to patients, counseling, and doorstep delivery of prescribed medication through trained village health workers contributed to treatment adherence and control status.
Keywords: NCD control, noncommunicable diseases, rural, treatment adherence
BACKGROUND
Globally, noncommunicable diseases (NCDs) account for about 41 million deaths per year (or 71% of all fatalities), including 14 million premature deaths. India has a higher NCD burden with NCDs accounting for roughly 65% of all deaths.[1] Nonadherence to medication is a core problem in NCD management. One in two patients fails to take the treatments as advised, and one-third of patients who take medicines stop their treatments earlier than required. About 22.4% of patients state ‘forgetting’ as their main reason for nonadherence.[2] It is essential to understand the various barriers to address this complex challenge. By helping improve adherence, we can help prevent or delay the onset of complications, reduce hospitalization risks, and decrease healthcare costs.[2]
Charutar Arogya Mandal (CAM), Karamsad, Gujarat, has been implementing Shree Krishna Hospital Initiative for Advancement of Rural and Social Health (SPARSH) as a community-based NCD care program since the year 2015. The program offers village-based screening of NCDs (hypertension and diabetes) among adults more than 30 years. Patients who get enrolled in SPARSH receive free medical consultation, delivery of prescribed generic medicine at doorstep, health education, and regular follow-up of their condition. These activities are carried out by a health team consisting of female village health workers, medical officers, pharmacists, and social workers. More details of the program have been reported elsewhere.[3,4,5] The stakeholders felt the need to understand the program’s effect on NCD control among enrolled patients and know how it worked. Hence, we carried out the present evaluation with the following questions: Is there an effect of the community-based SPARSH program on treatment adherence, treatment expenditure, and control among selected NCDs? What are the reasons for low treatment adherence to treatment from patients’ point of view?
MATERIAL AND METHODS
Setting
The present evaluation was carried out in 100 field practice villages of SPARSH in Anand, Kheda, and Panchamahal districts in Gujarat. The program was managed under Amrita Patel Centre for Public Health, Bhaikaka University, Karamsad.
Evaluation framework
One of the key strategies of the SPARSH program was focused on training of program staff and female village health workers and education of community members in NCD care for its control and prevention. Hence, we planned the present evaluation at level 3 of the Kirkpatrick framework, where there was a focus on outcome in the form of patients’ behavior change for treatment adherence in response to counseling and follow-up offered by the SPARSH team.[6] We modeled the present evaluation based on the pragmatism as we were centered to the problems that emerged from our own practice setting.[7]
Study design with a visual diagram
We used the explanatory type of mixed method design, where a major quantitative component (cross-sectional design) was followed by a minor qualitative (free list) component to answer (explain) the second research question in a single study design [Figure 1].
Figure 1.
Visual diagram of study design
Phase I: (QUAN)
Sample size and sampling
The sample size for the study was calculated by using the software Open Epi 6, where the desired precision considered was 5%, the expected treatment adherence and control status was 50%, and the design effect was 2, with 95% confidence interval. The calculated sample size was 769. After considering 10% of the nonrespondents, the final sample size was 845, which rounded off the nearest high figure of 850. This sample size was found to be adequate for multivariate analysis and yielded a power of 99.9%.
A two-stage cluster sampling technique was used to select a representative sample from the study area (project area). In stage 1, 30 clusters were selected by the probability proportional to size (PPS) technique. In the second stage, 29 participants (over 30 years) were selected per cluster by a random walk method[8] in each cluster. The random walk method was implemented in two steps: At the first step, the center of the villages was to be identified. Later, a pen was to get the direction for the initiation of sample selection. In the selected lane, the first household was selected by a lottery method and subsequent houses were selected till the sample size of 29 NCD patients known to have hypertension, diabetes, or both was met. Respondents with self-reported history of taking antihypertensive or antidiabetic medication were considered hypertensive or diabetic patients, respectively. If any house was found locked or there was unavailability of family members, the next house was selected. In the selected household, if there was more than one eligible participant, one participant was selected by the kish grid method.[9]
Data collection
After obtaining the informed consent, data were collected in face-to-face interviews with patients by using predesigned and pretested questionnaires. A trained female postgraduate in Community Medicine (first author), who was not part of the SPARSH program, collected the data. Information on sociodemographic details and treatment expenditure in the last year was collected. Socioeconomic status was assessed by using modified Dr. B.G. Prasad’s socioeconomic classification. Patients were asked about the duration of illness, history of tobacco, and alcohol consumption. Treatment adherence was measured using Morisky–Green–Levin (MGL) Adherence scale.[10] This scale is in the public domain and measures self-reported medication behavior among patients on 0–4 scores, where 0 is considered as ‘high adherence’ and 1–2 and 3–4 scores are considered as ‘medium adherence’ and ‘low adherence’, respectively. WHO – STEPS guideline was used for the measurement of height and weight.[11] Body Mass Index was classified as BMI values <18.5 as underweight, 18.5–22.9 as normal, 23–24.9 as overweight, and ≥25 as obesity.[12] Arterial blood pressure was measured via a digital blood pressure monitor, and blood glucose was measured using a glucometer. Three blood pressure measurements were taken 3 min apart, and the average of three readings was taken. A blood pressure of more than 140/90 mmHg was considered as uncontrol status among known hypertensive.[13] Random blood sugar of 200 mg/dl or higher was considered uncontrolled.[14,15] Enrolment in the SPARSH program meant the patients who agreed to take follow-up and prescribed medicine from the program team. If available, patients’ medical records were reviewed for more authentic information.
Data entry and analysis
The quantitative data were entered in Microsoft Excel and analyzed in STATA software. Based on existing literature, we developed a directed acyclic graph (DAG) as a causal diagram to understand the pathway and confounders using DAGitty v3.0 [Figure 2].[16] We wanted to test the hypothesis if enrolment in the SPARSH program improved the control of diabetes and hypertension among enrolled NCD patients. Control status (control of diabetes and hypertension) was our outcome variable, and enrolment in SPARSH was the exposure variable. Among the variables, DAG suggested that age, sex, education, socioeconomic status, occupation, and duration of illness were the confounders. It identified three causal pathways: 1) direct from SPARSH enrolment to NCD control status, 2) through adherence, and 3) through obesity [Figure 2]. Data fulfilled the assumptions of independence satisfactorily highlighted by the DAGitty using Chi-square test. We used path analysis to study the effect of SPARSH enrolment on control status among NCD patients and its causal pathway. Path analysis is a form of multiple regression analysis that is used to evaluate causal models and distinguish the overall causal effect into direct and indirect effects. Thus, path analysis helps understand the causal pathway between exposure and outcome, which is not possible in conventional regression analysis.[17]
Figure 2.
DAG with adjustment of SES and education
Phase II: (qual)
In a minor qualitative phase, we used a rapid ethnographic technique called free listing to understand the barriers to treatment adherence. We asked free list questions to all patients (n = 30) with low treatment adherence (a score of 3 to 4).[18] The primary stimulus question was 1) “What according to you are the reasons for poor adherence to treatment?”. The responses for each respondent were noted in the order in which it was mentioned by each of the participants. The data entry and analysis were done by using the Anthropac (4.98.1/X) software package. Smith’s S value was obtained for each item. Smith’s S values closer to 1 indicate the prominence of items across the lists. We used the reporting guidelines for mixed method research called Good Reporting of A Mixed Methods Study (GRAMMS) as a direction for reporting the present study.[19]
Ethical considerations
The study was approved by the Institutional Ethics Committee of Bhaikaka University, Karamsad (IEC/BU/139/Faculty/16/205/2022).
RESULTS
Among 850 surveyed patients, 420 (49.4%) were hypertensive, 179 (21.1%) had type 2 diabetes, and 251 (29.5%) had both hypertension and diabetes. Table 1 provides the sociodemographic details of the study participants.
Table 1.
Sociodemographic details of HTN and DM respondents [n=850]
| n (%) | |
|---|---|
| Age in years | |
| 30-60 | 416 (48.9) |
| More than 60 | 434 (51.1) |
| Sex | |
| Male | 328 (38.6) |
| Female | 522 (61.4) |
| Education status | |
| Illiterate | 104 (12.2) |
| Primary | 373 (43.9) |
| Secondary | 299 (35.2) |
| Graduate and above | 74 (8.7) |
| Occupation | |
| Involve less physical activity | 497 (58.5) |
| Involve more physical activity | 353 (41.5) |
| Religion | |
| Hindu | 800 (94.1) |
| Muslim | 45 (5.3) |
| Christian | 5 (0.6) |
| Socioeconomic status | |
| Class I – Upper class | 93 (10.9) |
| Class II – Upper middle class | 247 (29) |
| Class III – Middle class | 269 (31.7) |
| Class IV – Lower middle class | 153 (18) |
| Class V – Lower class | 88 (10.4) |
| Marital status | |
| Married | 557 (65.5) |
| Widowed + Widower + Divorced | 291 (34.3) |
| Unmarried | 2 (0.2) |
| Membership in groups | |
| SHG/Mahila/Sakhi Mandal/Bhajan Mandal | 19 (2.2) |
| None | 831 (97.8) |
| Current NCD Status | |
| Hypertension (HTN) | 420 (49.4) |
| Type-2 Diabetes (DM) | 179 (21.1) |
| Both | 251 (29.5) |
| Health insurance | |
| MA/Ayushman card | 522 (61.4) |
| Ashirwad and Krupa card | 95 (11.2) |
| LIC | 3 (0.3) |
| None | 230 (27.1) |
Among 202 patients enrolled in SPARSH, 182 (90.1%; CI: 84.9–93.7) had high adherence to treatment, 17 (8.4%; CI: 5.12–13.4) patients had medium adherence, and only three (1.5%; CI: 0.4–4.6) had low adherence to treatment. Noteworthily, 182 (90.1%; CI: 84.9–93.7) patients enrolled in the SPARSH program had their NCD condition under control status. There was significantly better adherence and control among patients enrolled in the SPARSH program as compared to those who were not enrolled (P < 0.01) [Table 2].
Table 2.
Status of treatment adherence and control status among SPARSH and non-SPARSH patients
| Variables | Patients enrolled in SPARSH (n=202) n (%; 95% CI) |
Patients not enrolled non-SPARSH (n=648) n (%; 95% CI) |
Test of significance |
|---|---|---|---|
| Adherence to treatment | |||
| High adherence (0) | 182 (90.1; 84.9-93.7) | 204 (31.5; 27.9-35.2) | 213.45, df=1, P<0.001 |
| Medium adherence (1-2) | 17 (8.4; 5.12-13.4) | 417 (64.3; 60.5-68.02) | 192.82, df=1, P<0.05 |
| Low adherence (3-4) | 3 (1.5; 0.4-4.6) | 27 (4.2; 2.8-6.1) | 3.25, df=1, P=0.07 |
| Control status | |||
| Yes | 182 (90.1; 84.9-93.7) | 457 (70.5; 66.8-73.9) | 31.62, df=1, P<0.01 |
| No | 20 (9.9; 6.3-15.1) | 191 (29.5; 26.0-33.2) |
The median direct expenditure for treatment of hypertension among patients enrolled in SPARSH was 480 INR (IQR: 276-770), which was 3300 INR (IQR: 1230-6000) among patients not enrolled in SPARSH. The median expenditure for treatment of diabetes among patients enrolled in SPARSH was found as 480 INR (IQR: 480-960), which was 3780 INR (IQR: 1780-7600) among patients not enrolled in SPARSH. The median expenditure for treatment of both hypertension and diabetes among patients enrolled in SPARSH was 970 INR (IQR: 804-1150), which was 9000 INR (IQR: 3940-13,030) among patients not enrolled in SPARSH [Table 3].
Table 3.
Treatment expenditure among the study participants (n=850)
| NCDs | Type of treatment expenditure in INR | SPARSH (n=202) Median (IQR) |
Non-SPARSH (n=648) Median (IQR) |
P # |
|---|---|---|---|---|
| Hypertension | Direct | 480 (276 – 770) | 3300 (1230 – 6000) | <0.001* |
| Indirect | 0 | 700 (300 – 1000) | NA | |
| Total | 480 (276 – 770) | 4176 (1630 – 6800) | <0.001* | |
| Diabetes | Direct | 480 (480 – 960) | 3780 (1780 – 7600) | <0.001* |
| Indirect | 0 | 700 (200 – 1700) | NA | |
| Total | 480 (480 – 960) | 4595 (2233 – 8400) | <0.001* | |
| Both hypertension and diabetes | Direct | 970 (804 – 1150) | 9000 (3940 – 13030) | <0.001* |
| Indirect | 0 | 1000 (600 – 2000) | NA | |
| Total | 970 (804 – 1150) | 10000 (5060 – 14530) | <0.001* |
Note: IQR – Interquartile Range, NA-Not Applicable, #based on Mann–Whitney U test, *Statistically significant, P<0.05
The effect of SPARSH enrolment on the control of hypertension and diabetes is adjusted for age, sex, occupation, socioeconomic status, and duration of illness. In addition to the direct effect of SPARSH enrolment on control, the effect is mediated through adherence, obesity, tobacco, and alcohol [Figure 2]. The final causal pathways of the effect of SPARSH enrolment on the control of hypertension and diabetes adjusted for the confounders is shown in Figure 3. The only significant path identified was through adherence. Enrolment in the SPARSH program increased adherence (beta = 0.903, P < 0.001), which led to control of hypertension and diabetes (beta = 0.191, P < 0.001). The total effect of SPARSH enrolment through adherence was 0.172. The other paths through obesity, tobacco, and alcohol including direct paths were not statistically significant [Table 4].
Figure 3.
Final causal diagram showing effects of SPARSH enrolment on control status of hypertension and diabetes
Table 4.
Path analysis showing causal pathways of SPARSH enrolment to control of hypertension and diabetes
| Variable | Coefficient | 95% Confidence Interval | P |
|---|---|---|---|
| SPARSH enrolment | |||
| Education | 0.006 | -0.001 to -0.014 | 0.072 |
| Sex | -0.117 | -0.178 to -0.056 | <0.001 |
| Occupation | -0.192 | -0.252 to -0.132 | <0.001 |
| Socioeconomic status | -0.091 | -0.115 to -0.066 | <0.001 |
| Duration of illness | -0.047 | -0.091 to -0.005 | 0.030 |
| Age | 0.001 | -0.002 to -0.003 | 0.756 |
| Control status | |||
| SPARSH enrolment | -0.019 | -0.088 to 0.050 | 0.593 |
| Education | 0.002 | -0.004 to 0.009 | 0.398 |
| Sex | -0.044 | -0.102 to 0.014 | 0,139 |
| Occupation | -0.023 | -0.081 to 0.034 | 0.437 |
| Socioeconomic status | -0.008 | -0.032 to 0.016 | 0.497 |
| Duration of illness | -0.071 | -0.111 to -0.029 | 0.001 |
| Age | -0.002 | -0.003 to 0.001 | 0.098 |
| Adherence | 0.192 | 0.162 to 0.221 | <0.001 |
| Obesity | 0.001 | -0.051 to 0.052 | 0.968 |
| Tobacco | -0.001 | -0.054 to 0.052 | 0.971 |
| Alcohol | -0.034 | -0.119 to 0.051 | 0.429 |
| Obesity | |||
| SPARSH enrolment | -0.025 | -0.102 to 0.051 | 0.507 |
| Adherence | |||
| SPARSH enrolment | 0.903 | 0.765 to 1.041 | <0.001 |
| Tobacco | |||
| SPARSH enrolment | 0.011 | -0.067 to 0.889 | 0.783 |
| Alcohol | |||
| SPARSH enrolment | -0.004 | -0.053 to -0.044 | 0.870 |
LR test of model vs. saturated: Chi2 (29) = 46.09, Prob>chi2=0.0230, RMSEA=0.026
The top three reasons for low adherence were 1) forgetting to take the medicine due to a busy schedule, 2) lack of money for medicine purchase, and 3) better feeling, so felt not necessary to take medicine [Table 5].
Table 5.
Reasons for low treatment adherence among patients with low adherence [n=30]
| Items | Frequency (%) | Average Rank | Salience |
|---|---|---|---|
| Due to my busy schedule forget to take my medicine | 63.3 | 1.74 | 0.458 |
| Lack of money for medicine purchase | 53.3 | 2 | 0.333 |
| Feel better, so no need to take medicine | 50 | 1.47 | 0.419 |
| Fear that medicine will cause health issues | 33.3 | 2.4 | 0.183 |
| Due to lack of transportation and laziness problems | 20 | 1.83 | 0.153 |
| Already taking medicine for other health problems. So not taking medicine for NCDs | 6.7 | 1.5 | 0.056 |
| Due to a lack of medicine stock in Govt hospital | 6.7 | 2 | 0.044 |
| Not feeling good after taking medicine | 3.3 | 3 | 0.011 |
| Taking some home remedies | 3.3 | 1 | 0.033 |
| Lack of knowledge | 3.3 | 2 | 0.025 |
DISCUSSION
Enrolment in the SPARSH program increased treatment adherence among NCD patients, which led to control of hypertension and diabetes. Among patients enrolled in SPARSH, 90% had high adherence to drug treatment. The median treatment expenditure among patients enrolled in the SPARSH program was significantly lower than those who were taking treatment elsewhere. As prescribed medicines were delivered at the doorstep, the enrolled patients in SPARSH had no indirect expenditures and had to bear only the actual cost of medicine. Reasons for low adherence were related to patients’ forgetfulness to take medicine as per schedule, lack of money to buy the medicine, and sense of feeling better. Regular follow-up visits by village health workers and door-step delivery of medicine seemed to have addressed some of these problems among patients enrolled in the SPARSH program.
A previous community-based study in Gujarat on medication adherence among NCD patients showed that 41% of patients with hypertension and 37% of patients with diabetes had high adherence to treatment.[20] K. Yuvaraj et al.[21] did a facility-based cross-sectional study in rural Puducherry and found that one-third participants had low adherence to treatment. We found that only 1.5% patients enrolled in SPARSH had low treatment adherence. The top three reasons stated by NCD patients for low adherence were 1) forgetting to take the medicines due to busy schedule, 2) lack of money to purchase medicines, and 3) feeling better, so felt not necessary to take medicines. Hence, NCD patients need a regular supply of medicine at an affordable cost, motivation through follow-up visits, and education on the asymptomatic nature of illness during the early phase of the progression.
It was sequential mixed methods evaluation that helped us to get the complete picture of the situation in a single design where qualitative findings offered additional information to improve the program further. We could thus test if the SPARSH program was effective among enrolled NCD patients and explore the context-specific reasons for “low adherence”. The survey was done on a large representative sample in a wider geographical area. The first author who collected the data was not part of the SPARSH team, which might have reduced the desirability bias. Investigator triangulation from different departments and institutes might have reduced the bias in analysis and interpretations. We have used a DAG and path analysis to improve the rigor of multivariable analysis.
The limitations of the present study should be kept in mind. It was a context-specific evaluation, and its findings may be useful to those who work in similar practice settings. Due to practical feasibility at the field level, we relied on measurement of random blood sugar for deciding the control status of diabetes. Even though two-third of the patients had Ayushman card and there was provision of free supply of medicine under the national health program, treatment expenditure among patients not enrolled in SPARSH was high. This finding needs more follow-up and explanatory research. Finally, as it was a cross-sectional survey, the temporality of the association between exposure and outcome is hard to demonstrate, making it difficult to infer causality.
CONCLUSIONS
To conclude, enrolment in the SPARSH program increased treatment adherence, which led to control of hypertension and diabetes among NCD patients. Home visits to patients and doorstep delivery of prescribed medication through trained village health workers improved their treatment adherence and control status.
Conflicts of interest
There are no conflicts of interest.
Acknowledgements
We acknowledge the financial grant from the Sir Dhorabji TATA Trust and GMM Pfaudler (SDTT/MUM/HEA/CAM/2017-2018/0054) Limited for SPARSH programme. We acknowledge Dr. Vinayagamoorthy, Assistant Professor, AIIMS Deoghar for his technical help.
Funding Statement
Nil.
REFERENCES
- 1.Noncommunicable diseases. [[Last accessed on 2023 Feb 14]]. Available from: https://www.who.int/news-room/fact-/detail/noncommunicable-diseases .
- 2.Improving medicine adherence to fight India’s NCD burden | Mint. [[Last accessed on 2023 Feb 15]]. Available from: https://www.livemint.com/science/health/improving-medicine-adherence-to-fight-india-s-ncd-burden-11667463837824.html .
- 3.Satish P, Khetan A, Raithatha S, Bhende P, Josephson R. Standardizing hypertension management in a primary care setting in India through a protocol based model. Indian Heart J. 2019;71:375–80. doi: 10.1016/j.ihj.2019.11.257. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kumar D, Raithatha SJ, Gupta S, Raj R, Kharod N. Burden of self-reported noncommunicable diseases in 26 villages of Anand District of Gujarat, India. Int J Chronic Dis. 2015;2015:1–6. doi: 10.1155/2015/260143. doi: 10.1155/2015/260143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Roy K, Das N, Roy SA. Strengthening primary health care in rural western India-team based approach. Int J Res Med Sci. 2023;11:1303–8. [Google Scholar]
- 6.Kirkpatrick D, Kirkpatrick J. 3rd. Oakland: Berrett-Koehler; 2006. Evaluating Training Programs: The Four Levels. [Google Scholar]
- 7.Creswell JW, Plano Clark VL. 2nd. New Delhi: SAGE Publications; 2011. Designing, and Conducting Mixed Methods Research. [Google Scholar]
- 8.Alvi Munich Personal RePEc Archive A Manual for Selecting Sampling Techniques in Research. 2016. [[Last Accessed on 2024 Jul 04]]. Available from: https://mpra.ub.uni-muenchen.de/70218/1/MPRA_paper_70218.pdf .
- 9.Kish Grid: What It Is and How to Use It-Statistics How To. [[Last accessed on 2022 Dec 24]]. Available from: https://www.statisticshowto.com/kish-grid/ .
- 10.Morisky DE, Green LW, Levine DM. Concurrent and predictive validity of a self-reported measure of medication adherence. Med Care. 1986;24:67–74. doi: 10.1097/00005650-198601000-00007. [DOI] [PubMed] [Google Scholar]
- 11.Standard STEPS instrument. [[Last accessed on 2022 Dec 29]]. Available from: https://www.who.int/teams/noncommunicable-diseases/surveillance/systems-tools/steps/instrument .
- 12.Verma M, Rajput M, Kishore K, Kathirvel S. Asian BMI criteria are better than WHO criteria in predicting Hypertension: A cross-sectional study from rural India. J Family Med Prim Care. 2019;8:2095–100. doi: 10.4103/jfmpc.jfmpc_257_19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Armstrong C. JNC 8 guidelines for the management of hypertension in adults. Am Fam Physician. 2014;90:503–4. [PubMed] [Google Scholar]
- 14.Susairaj P, Snehalatha C, Raghavan A, Nanditha A, Vinitha R, Satheesh K, et al. Cut-off value of random blood glucose among Asian Indians for preliminary screening of persons with prediabetes and undetected type 2 diabetes defined by the glycosylated haemoglobin criteria. J Diabetes Clin Res. 2019;1:53–8. doi: 10.33696/diabetes.1.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Diabetes Mellitus: When to suspect/recognize. National Health Mission. [[Last accessed on 2024 June 04]]. Available from: https://www.nhm.gov.in/images/pdf/guidelines/nrhm-guidelines/stg/diabetes-mellitus.pdf .
- 16.Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology. 1999;10:37–48. [PubMed] [Google Scholar]
- 17.Streiner DL. Finding our way: An introduction to path analysis. Can J Psychiatry. 2005;50:115–22. doi: 10.1177/070674370505000207. [DOI] [PubMed] [Google Scholar]
- 18.Dongre AR, Deshmukh PR. 1st. Chennai: Notion Press; 2021. Practical Guide: Qualitative Methods in Health and Educational Research. [Google Scholar]
- 19.Gugsa F, Karmarkar E, Cheyne A, Yamey G. Good reporting of a mixed methods study (GRAMMS) newspaper coverage of maternal health in Bangladesh, Rwanda, and South Africa: A quantitative and qualitative content analysis. BMJ Open. 2016;6:e008837. doi: 10.1136/bmjopen-2015-008837. doi: 10.1136/bmjopen-2015-008837. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Lakshmi N, Algotar P, Bhagyalakshmi A. Community based study on medication adherence and its determinants among people with Noncommunicable diseases in Ahmedabad. Int J Multidiscip Res Dev. 2019;6:1–4. [Google Scholar]
- 21.Yuvaraj K, Gokul S, Sivaranjini K, Manikandanesan S, Murali S, Surendran G, et al. Prevalence of medication adherence and its associated factors among patients with noncommunicable disease in rural Puducherry, South India – A facility-based cross-sectional study. J Fam Med Prim Care. 2019;8:701. doi: 10.4103/jfmpc.jfmpc_350_18. [DOI] [PMC free article] [PubMed] [Google Scholar]