Study design and setting
A cross-sectional interview survey was conducted in all 17 urban primary healthcare centres in Bhubaneswar, the capital city of Odisha with a population of 900,000 inhabitants. [12] According to the National Sample Survey Office’s 71st round on social consumption of health, about 72% of outpatient care in Odisha is provided by public healthcare professionals. [13] The public health care system has a three-tier structure (primary, secondary and tertiary levels). Primary Health Care Centres are involved in delivering primary care while district hospitals and sub-divisional hospitals provide secondary care. Tertiary health care is provided by medical college hospitals.
Study participants
Patients attending a primary healthcare center between September 2014 and February 2015 who had been diagnosed by a physician with T2DM for more than six months according to their personal medical record were eligible to be included in the study. Since the consultation time is limited in the healthcare centers and every interview took 20-30 minutes and only one interviewer was available per center, every third eligible diabetes patient was invited to participate to make the study feasible. The inclusion criterion of diabetes duration of at least six months was applied because we needed information about healthcare utilisation for diabetes. Patients too ill to participate or with emergency health conditions were excluded from the study. Anonymised details of all patients excluded (age, gender, reason for exclusion) were recorded to compare the characteristics of the participants with the non-participants.
Measurements
The participating patients were interviewed in a separate private room using a predesigned and pretested questionnaire, Diabetes Co-morbidity Evaluation Tool in Primary Care (DCET- PC). The DCET-PC is derived from “Multimorbidity Assessment Questionnaire for Primary Care”, a validated questionnaire which was pretested and the feedback used to adapt the questionnaire for our study. [14] Two graduate nurses trained in patient history-taking and interview techniques carried out the interviews, and 10% of the interviews were done in the presence of the first author. The DCET-PC (Appendix 1) included questions about the existence of comorbid conditions, eliciting information on whether the patient had any of the16 listed chronic conditions, and socio-demographic details, i.e. age, sex, residence (rural, semi-urban, urban), ethnicity (general, scheduled caste and tribe, other backward classes), religion (Hindu, Muslim, Christian, others), educational level (no education, primary level, secondary, graduate and above), marital status (single, married), annual family income (categorised into five quintiles) and household status (above poverty line, below poverty line). The details of development and domains of the DCET-PC questionnaire were described in our previous paper.[1]
We estimated comorbidity in three ways: 1) presence or absence of any comorbidity, which was further categorised into 2) the number of comorbid conditions (zero, one, two, three, four or more chronic conditions), and 3) the presence of any one of the 16 chronic conditions in our study in one individual patient. Healthcare utilisation was operationalized as the reported number of visits to any healthcare facility in the last six months for any reason. Expenditure was measured in Indian Rupees (INR) by asking about expenses incurred in the last six months separately for outpatient consultation fees, medicines (for DM and other diseases separately), travelling to those healthcare facilities, and diagnostic tests (for DM and other diseases separately). Total out-of-pocket expenditure was defined as the sum of these costs.
Analysis
To estimate the healthcare utilization, due to the skewed natured of the data, median (interquartile ranges) number of visits done by the patient to any healthcare facility during last six months were calculated. Healthcare utilization and out-of-pocket expenditure were further described across the number of comorbid conditions and the prevalence of leading comorbidities. Bivariate comparison was performed using a Kruskal Wallis test for quantitative data (on the basis of median values) and a chi-square test for categorical data. Furthermore, we calculated the median and interquartile ranges of out-of-pocket expenditure by comorbidity status (Yes/No). The difference in mean out-of-pocket expenditure and healthcare utilization across the comorbidity groups was tested using Wilcoxson signed rank test.
Poisson regression model in multilevel mixed effects methods was used with two levels (health center and patient) for multivariate analysis to assess the independent contribution of comorbidity on healthcare utilization and out-of-pocket expenditure. The collinearity between the variables was tested before including them in multivariate analysis. Adjusted incidence rate ratio was calculated for each predictor for estimating health care utilization and expenditure. A p-value of <0.05 was considered statistically significant. Analyses were performed in STATA Corp-12 Tx.