The fourth round of the District Level Household and Facility Survey (DLHS-4) is the most comprehensive and latest nationwide health facility assessment conducted by the government of India with coverage across 26 states and 3 union territories with detailed district-level data. The survey was conducted between 2012–2014. Facility data were collected through four separate modules (including questionnaire, physical inspection, and assessing registers) designed for each facility type. Facility data were available for 1,540 District and Sub-District Hospitals (DH), 4,810 Community Health Centers (CHCs), 8,540 Primary Health Centers (PHCs), and 18,367 Sub Centers (SCs). In 377 districts, we analyzed DH facilities meeting the Indian Public Health Standards (IPHS) minimum bed criterion of 101 beds; in 171 districts where no facility met this criterion, we analyzed sub-divisional/sub-district hospitals (minimum of 31 beds). Data from two states (Gujarat and Jammu and Kashmir) and four union territories (Dadra and Nagar Haveli, Daman and Diu, Delhi and Lakshadweep) were not available.
To characterize broader community context at the district-level, we combined data from the household survey of the DLHS-4, the Annual Health Survey (AHS),19–21 and the fourth round of the National Family Health Survey (NFHS-4).22 The DLHS-4 household survey was conducted in 2012-14 in all states and union territories in India except the for the states of Bihar, Chhattisgarh, Jharkhand, Madhya Pradesh, Odisha, Rajasthan, Uttar Pradesh, Uttarakhand and Assam. States that were not covered in the DLHS household survey were instead covered through the government’s AHS, conducted in 2012 − 13.21 Together, the DLHS-4 and AHS provide coverage of districts across all Indian states in 2012–2014. NFHS-4 was conducted in 2015 − 1622 and provides data and households with district linkages. NFHS employed a multi-stage stratified sampling scheme and were designed to be representative at the state and national levels. Strata were defined by urban-rural setting and the primary sampling units (PSU) were villages in the rural stratum and wards in the urban stratum. NFHS provides data for all of the districts under study, with the advantage of having identical measures across all districts.
The CICI framework was developed to provide guidance on the interacting dimensions of context, implementation, and settings that may impact the successful delivery of complex interventions.18 Critically, the framework fills the gap in how upstream contextual factors beyond the organizational context may impact the implementation of a complex intervention with community-facing components. CICI considers 7 domains of context—geographical, epidemiological, socio-cultural, socio-economic, ethical, legal, political. Here, we investigate the interplay among the CICI described political (healthcare infrastructure), epidemiological (measures of blood pressure, body mass index, older population), socioeconomic (aggregate community wealth), and geographical (region, urbanicity) domains. Although socio-cultural, ethical, and legal aspects warrant consideration, they are not easily quantifiable and are not included in the present study.
The CICI places healthcare infrastructure within the political domain because it is dependent on healthcare financing and regulations. As healthcare infrastructure is the sine qua non of medical management of blood pressure, this is the focal domain of this study. The public healthcare system in India is organized as a hierarchy of four tiers of facilities that include: sub-centres at the village-level (SC; lowest levels of skilled personnel and resources; tasked with blood pressure screening), primary health centres that serve several villages (PHCs; contain a physician; tasked with blood pressure diagnosis and providing basic medical treatment), community health centres that serve the administrative unit known as a block (CHCs; include NCD clinics and are central to the integration of NCD care into primary care), and district hospitals that serve an entire district (DH; include higher levels of trained personnel and more sophisticated infrastructure). This model was designed to broaden coverage within existing resources and with the aim of progressive referral from lower to higher levels of health care depending of the need of the individual patient and the availability of resources (skilled human resources, infrastructure and services).
Healthcare system infrastructure indicators were developed based on a combination of the Indian Public Health Standards (2012 revision).23 We used the DLHS facility data to separately quantify the availability of diagnostics (blood pressure instrument), antihypertensive medication, potential staffing (medical officer, staff nurse, pharmacist, and community health worker), and specific infrastructure needed to support digital health initiatives (power supply, computer, internet connection). Indicators were contingent on the facility tier (DH, CHC, PHC, SCs). For example, at the district hospital level, a medical officer, public health nurse, and a pharmacist are designated as essential staff, whereas in health sub-centres (lowest level), only a female health worker is designated as essential. We only included personnel who, from our experience, are likely to be involved in hypertension care. The availability of indicators across tier of the healthcare system hierarchy is described in Supplemental Table 1.
Epidemiologic, socioeconomic and geographical context
To assess the alignment of healthcare infrastructure with broader contextual characteristics, we also considered measures of the epidemiologic, socioeconomic, and geographical context. Such contextual indicators may have a bearing on allocation of health system resources. For example, health system infrastructure relevant to blood pressure may be directed to communities with greater epidemiologic burden of hypertension and its risk factors. Alternatively, it may be the case that wealthier communities may be the recipients of larger public investments in health.
Districts were characterized for epidemiologic and geographical context by aggregating household and individual data from the combined DLHS and AHS datasets. For epidemiologic indicators, we computed the mean systolic blood pressure and the mean body mass index by district. Geographical context was defined by region as well as the proportion of the population that resides in an urban area. Districts were classified as belonging to one of six regions—north (RJ, UT, HR, PJ, HP, CH), west (MH, GA), south (TN, AP, AN, KL, KA, PY), central (CG, MP, UP), east (BR, JH, OR, WB) and northeast (AP, AS, MN, MZ, ML, NL, SK, TR)—according to the Indian Census. For socioeconomic context, we assigned each district the mean of measure of household wealth, a standardized asset-based measure of relative household wealth (mean = 0 and SD = 1) computed separately for urban and rural households. District-level wealth scores were derived from NFHS-4, for a measure of relative wealth comparable across all districts.
Districts were the unit of analysis for this study. All statistical analyses were conducted in SAS v9.4 (SAS Institute; Cary, NC). Missing data was assumed to indicate the absence of that the factor; for example, facilities lacking missing data on internet availability were assumed to have no internet.
District infrastructure indicators were first computed for all health facilities separately. Health facilities were scored 1 for the presence or 0 for the absence of each indicator available for its facility type. A composite variable was created for each facility and was coded as 1 if a facility reporting having all required indicators (diagnostics, medication, potential staffing, and IT infrastructure). District-level infrastructure scores were computed as the mean of health facility context indicators for all facilities within a district. For example, the mean of the overall CHC facility composite score reflects the proportion of CHCs in a particular district that met all of the infrastructure criteria listed above.
We conducted a descriptive analysis of all indicators and composite scores and estimated means and 95% confidence intervals (CIs) for the nation and by region. We computed the Spearman correlation among all of the contextual indicators, including the mean composite healthcare system infrastructure availability, at the district level. We further conducted logistic regression analysis to estimate the association between epidemiologic and geographical context (exposures) and the composite measure of healthcare system potential to support digital health interventions for blood pressure management (outcome) at the facility level. Models were implemented using generalized estimating equations to account for clustering of districts within states and were estimated separately for each facility tier (i.e., DH, CHC, PHC, and SC).
Finally, because the DLHS-4 was fielded in 2012, we conducted a sensitivity analysis to assess the correlation between total staffing of essential personnel at CHCs in 2012 and 2018 at the state level. Due to data limitations, only state level comparisons of staffing were possible and we focus on the CHC due to its emphasized role in NCD care.