The QIDS Study

Study Design and Sampling Scheme

The study utilizes an experimental design where the two interventions plus a control are randomized to matched blocks of hospital districts. This study design facilitates the measure and analysis of potential intervention impacts. Thirty hospitals in the Central Philippines were chosen using a blocked design from among the government’™s 64 local government units (LGU) that were identified for reform under the Health Sector Reform Agenda. The 30 selected LGUs were matched in blocks of 3 based on the population and system characteristics. These characteristics include average income, labor force participation rate, infant mortality rate, functional literacy, and percentage of the population with insurance. To reduce within block variation for each population characteristic, an attempt was made to match locations to within 10 to 25% of existing variation between administrative regions. The key system characteristics used for matching the LGU blocks include number of specialists (< or equal to 20%), proximity to Manila, and the existence of a regional or specialist hospital (both < or equal to 50 km). Once each matched block of three LGUs was made, sites were randomly selected: one intervention site for expanded insurance, one for bonus payments, and one control site. Nine of these matched blocks are located within the Visayas and within the same province. The tenth matched block consists of 3 island hospitals in Siquijor, Biliran, and Camiguin. All study sites signified their intent to be part of the study through Memoranda of Agreements (MOA) with the Study Team. About one million households reside in the catchment areas of the 30 QIDS hospitals.

The QIDS' sample frame and data collection figure below reveals the longitudinal and multi-level components of the study. During each round, data is collected from 30 public hospital facilities, 10 per the 3 intervention groups and 3 randomly selected public providers from each hospital, totaling 90 public doctors in the sample at each round. In addition, 2 private doctors from the same catchment area and who attend to pediatric cases are randomly selected, totaling 60 private physicians in each round. In total, there are 150 physicians who complete clinical vignettes and physician surveys during each round.

Additionally, during each survey round, 100 children per hospital are sampled for the patient exit survey, which collects data on children between the ages of 6 months and 6 years as they are discharged from the facility; 30 of these children are those who are admitted to the hospital for diarrhea and 30 others are those who are admitted for pneumonia, our two tracer conditions. In total, 3,000 children per round are administered exit surveys, 1,800 of which have one of our two tracer conditions and will later be administered the household survey.

Four to ten weeks after the exit survey, follow-home household surveys are conducted on the 1,800 patients with tracer conditions. Two years later, in a subsequent round of data collection, household panel surveys are administered to these followed-home children. In other words, patients are administered a series of household surveys linked to the initial health insult starting 4-10 weeks after the exit survey, and then every 2 years thereafter.


Click for a larger image

So that policy effects can be quantified, measures of children’s health including cognitive development and the quality-of-care they receive are collected as a parallel process to the policy implementation. Apart from the usual anthropometric measures (height and weight), blood samples are drawn to generate other physiologic measures of health such as blood levels of hemoglobin, lead, folate, and C-reactive protein. Quality of care measures are collected on both demand and supply sides. On the patient’s end, health care outcomes are observed by conducting a follow-up interview of patients in their homes. On the provider’s end, the quality of clinical practice is assessed by giving doctors vignettes or paper cases of children with typical illnesses.

So that observed outcomes can be properly attributed to policy interventions, it is necessary to observe policy outcomes before and after the policy interventions are introduced in sites where policy interventions are in place or not. Moreover, since data from patients is gathered through patient exit surveys in hospitals, possible selection biases are corrected by also collecting data from a random sample of households as well.

Back to top

Interventions

To determine the effects of the two health care reforms in the Philippines on child health outcomes including cognitive development, the study pursues two main hypotheses:

A. Health status in children, including cognitive development, will be improved when their insurance benefits expand to 100% (Access Intervention).
B. Health outcomes in children, including cognitive development, will improve when health care providers are given performance-based payments resulting in improved quality of clinical practice (Bonus Intervention).

The Access Intervention refers to expanded and assured financial access to health care coverage (“zero co-pay”) for confinements of children five years old and younger (i.e. under 6). To facilitate the zero co-pay intervention, admissions of eligible dependents will automatically be classified as intensive cases unless the case is already of higher classification.

Under the Bonus Intervention, hospitals that have been found to meet certain pre-determined quality standards are eligible for bonus payments. These qualified hospitals are now allowed to make insurance claims for professional fees at the specialist rate. Eligibility for bonuses is determined using a quality metric which combines average vignette scores of randomly selected physicians in a hospital, facility case load, and average patient satisfaction. Each quarter, the QIDS quality metric is computed for each hospital assigned to this intervention. Hospitals that qualify are given a bonus that is payable to the Chief of Hospital, who in turn, distributes these payments to the hospital (medical and non-medical) staff.

Back to top