This study was part of a prospective mixed methods evaluation of the impact of accreditation on hospital quality, patient experience and organisational culture conducted in a large, publicly funded, university teaching hospital in Hong Kong. The hospital’s accreditation process began with a gap analysis based on ACHS standards and subsequent quality improvement initiatives to address identified gaps as described in detail elsewhere [18]. The hospital’s quality improvement activities included efforts to address hospital-wide issues, such as improving coordination, reporting and integration as well as specific department and procedure-level gaps. The current paper presents findings from the patient experience survey evaluating the effect of accreditation on care experience.
Study subjects
Acute care inpatients aged 18 to 80 were recruited to participate in the survey on the second day of their hospital admission at three time points corresponding with accreditation: nine months pre-accreditation (T1) as baseline, three (T2) and 15 (T3) months post-accreditation. Hospital admission and discharge staff identified eligible patients from the daily hospital admission records and prepared rosters for the research team. Patients were excluded if they were admitted to the ICU, Accident and Emergency observation, isolation, labour, private, psychiatric or custodial wards, had a psychiatric diagnosis, were in poor physical status, or were unable to communicate in Cantonese, Mandarin or English. Ward nursing staff confirmed patient eligibility.
Survey instrument
The widely used and internationally validated Picker Patient Experience Questionnaire-15 (PPE-15) was selected to evaluate participants’ experience with their recent inpatient episode. The PPE-15 measures seven aspects of inpatient experience: information and education, coordination of care, physical comfort, emotional support, respect for patient preferences, involvement of family and friends and community and transition [19]. The PPE-15 was used for the first examination of self-reported inpatient experience in Hong Kong, conducted as part of the 2005 Thematic Household Survey [3]. The 2005 data is used as a reference for this study.
Seven additional items collected patient demographic information, including education level, marital status, smoking and drinking history, self-perceived health, private health insurance coverage and medical benefits. The questionnaire was administered in Cantonese, English or Mandarin based on patient preference. Survey items were translated and back-translated by research staff and pilot tested to ensure accuracy and comprehension.
Procedure
The same procedure was followed for each of the three data collection periods in January-March 2010, 2011, and 2012. Research staff approached eligible patients in the ward and invited them to participate. Upon agreement, research staff obtained patient consent, contact telephone numbers and preferred time for follow up phone call. Trained research staff then contacted recruited patients one-week post discharge via telephone and if unable to reach, conducted up to five additional attempts at various times of day to increase the response rate. Response rate was calculated as the number of survey respondents over the number of patients who consented to participate and were contacted by telephone.
A unique study identification number (USI) was generated for recruited patients and linked to their corresponding Hospital Number (HN). HN is a per case (admission) hospital identification number. For each recruited patient, the Admission and Discharge office provided admission ward, discharge diagnoses (ICD 9 CM; up to 15), total number of hospital admissions in the preceding year, and length of stay information. All data analysis used de-identified data. All identifying information was excluded prior to analysis by employing USI numbers as the sole form of identification in the dataset. To ensure there was no risk of personal data being identified by name, all relevant information was encrypted and stored in a separate file, with the master linking file kept in a data safe.
Scoring and Analysis
Based on the previously validated PPE scoring scheme, we coded each item dichotomously to indicate the presence or absence of a problem (See Figure 1 for sample questions and scoring). “Problems" are aspects of the health care experience patients indicated could be improved. Summary scores were calculated as the ‘number of items identifying a problem’ over the ‘number of items answered by the patient’ on a scale of zero to 100, with zero indicating no problem and 100 indicating many problems. PPE-15 summary, domain and item scores were calculated for each collection period (T1, T2 and T3).
INSERT FIGURE 1
We first compared baseline (T1) PPE-15 scores for the study hospital to the Hong Kong public hospital scores obtained through the 2005 Thematic Household Survey (THS) [3]. The THS is a series of regular, repeated cross-sectional household surveys sampling the entire land-based Hong Kong population and covering a wide spectrum of social issues. Each survey usually focuses on two substantive policy subjects. The 2005 THS was the first to include patient experience as one of the survey sections, specifically inquiring about the most recent hospital admission in the 12 months prior to the survey. Hence, the survey items were only relevant for respondents with a recent admission and the recall period was for the entire year rather than one week as in our study data. We used t-tests to compare our hospital baseline scores to the 2005 Hong Kong average for public hospitals.
We then used ANOVA to evaluate differences in PPE scores across the three cross-sections with the Bonferroni post-hoc test for pairwise comparisons. Finally, we used multiple linear regression analysis with time period as the predictor variable and patient experience scores as the outcome variables to examine the effect of accreditation over time. Covariates included patient age, gender, self-reported education level, marital status, self-reported health status, smoking and alcohol use, insurance status, medical benefit status, length of stay, prior admissions and number of comorbidities. We also conducted sensitivity analyses for patients with lengths of stay (>4 days) or number of prior admissions (>1 prior admission) above the median since the ranges varied widely. All data analyses were conducted in STATA 13. A priori significance level of 0.05 was used for all statistical tests.
The study was approved by the Institutional Review Board of the institution and study hospital involved.