Step 2: Exit surveys
The goals of the exit survey were: (i) to document costs of
hospitalization (both medical and non-medical) at private and
public facilities, and in urban and rural areas; (ii) to explore the
utilization of different coping strategies; and (iii) to identify
poor households who could be interviewed, in-depth, in the
final step of field-work.
Eight-hundred respondents were sought, with equal numbers
in urban and rural areas, and equal numbers using the public
and private hospitals that were most commonly mentioned in
the focus-group discussions. In rural areas, respondents had to
be resident in the three sub-districts included in Step 1. Urban
respondents had to be resident in Vadodara city. In both urban
and rural areas, hospitals were purposively selected based on
frequency of use reported by respondents in the FGDs. The
rural hospitals tended to be much smaller than the urban
facilities; hospitals had to have a minimum of 15 inpatient beds
in order to be included in the study.
One hundred exit surveys were conducted at each of four
urban hospitals (two public and two private). Given the smaller
size of rural hospitals, interviews had to be conducted at six
facilities (three public and three private), with 65–70 respondents
per hospital. Potential respondents were identified by
having hospital administrators provide a list of patients to be
Figure 1 Mechanisms for coping with financial shocks. Source: World Bank (2001)
328 HEALTH POLICY AND PLANNING
Downloaded from http://heapol.oxfordjournals.org/ by guest on June 19, 2012
discharged on the day of exit interviews. Exit interviews were
restricted to those hospitalized in general wards (thus excluding
those who paid extra—both at public and private facilities—to
stay in private rooms). Patients (and their families) were
approached for interviews immediately after they made their
payments and had received their discharge cards. In order to be
included in the exit surveys, respondents had to be: (i) older
than 18 years of age; (ii) hospitalized for more than 24 hours;
and (iii) resident in the corresponding area (either urban or
rural) at the time of the interview (for example, residents of a
rural village hospitalized in urban facilities were excluded from
the study).
Interviews were conducted inside hospital premises by RJ and
three trained investigators. Data were collected using an
interview schedule which was filled out by the interviewer.
The following data were collected:
Place of residence and place of origin;
Details as to when they moved to their current place of
residence;
Cause of hospitalization;
Expenditures on hospitalization, with breakdown by type of
costs, e.g. medicines, doctors’ fees, etc;
Indicators of socio-economic status.
In most cases the patient was interviewed (generally with
their accompanying family present). In those cases where
patients were unable to respond (for example, if the patient
remained ill or unresponsive at the time of discharge) we
interviewed an accompanying household member instead. As
anticipated, 800 exit interviews were conducted. In no case did
potential respondents refuse to participate in the interviews.
Data were double-entered into an Excel database, and
cross-checked for any inconsistencies. Analyses were conducted
using the statistical software STATA. As a proxy for wealth, we
constructed a socio-economic status (SES) index based on
household assets and utilities, allowing the weights of these
assets to be determined by principal components analysis (PCA)
(Filmer and Pritchett 2001). All 26 assets and utilities variables
from the survey were retained in the index (see Appendix 1)
and weighted based on PCA. Twenty-one categorical variables
were converted to dichotomous variables as this provided for
greater discrimination amongst poorer households. Ultimately,
the index comprised 25 dichotomous variables and one
continuous variable (number of rooms). The index was
validated by examining the likelihood of ownership of specific
assets (or utilities) by decile. For example, it can be seen that
no respondent below the 50th percentile reported owning a
refrigerator, compared with 65% of respondents in the wealthiest
decile (Appendix 1). Respondents were grouped by quintile
or decile; in both cases the 1st was the poorest.