1 / 21

A propensity score approach to comparing medical costs between hospital districts

A propensity score approach to comparing medical costs between hospital districts. Hailuoto workshop 31. 5. 2005 Antti Liski. The study. a part of Reijo Sund’s (STAKES) ”Utilisation of routinely collected register data in health system performance assessment” –project

jon
Download Presentation

A propensity score approach to comparing medical costs between hospital districts

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. A propensity score approach to comparing medical costs between hospital districts A propensity score approach to comparing medical costs between hospital districts Hailuoto workshop 31. 5. 2005 Antti Liski

  2. A propensity score approach to comparing medical costs between hospital districts The study • a part of Reijo Sund’s (STAKES) ”Utilisation of routinely collected register data in health system performance assessment” –project • part of a bigger project led by Pekka Rissanen and funded by the Academy of Finland • my master’s thesis

  3. A propensity score approach to comparing medical costs between hospital districts The aim of this study • to compare the costs of hip fracture patients between care districts • there is indication that the prolonged in-hospital operative delay has an effect on mortality (Sund & Liski) • does the prolonged in-hospital operative delay have an effect also on the costs

  4. A propensity score approach to comparing medical costs between hospital districts Hip fractures • common with older people • lead to lower quality of life • usually surgically operated • the operation is expensive

  5. A propensity score approach to comparing medical costs between hospital districts The data • collected from the Finnish Health Care Register, Finnish Hospital Discharge Register, Finnish Health and Social Welfare Care Register, the data warehouse of the Finnish Hospital Benchmarking project and the National Causes of Death Register • the final data set consisted of 16881 hip fracture patients in 20 hospital districts in 1998-2001 that hadn’t had a hip fracture during the preceding 10 years

  6. A propensity score approach to comparing medical costs between hospital districts The data… • only surgically operated patients • three types of hip fractures • the costs are based on Diagnosis Related Costs (DRG) classification • every patient has the costs from year 2001 • the follow-up period was one year

  7. A propensity score approach to comparing medical costs between hospital districts Problems in comparing the costs of hospital districts • heterogeneity of patients in different hospital districts • censoring caused by deaths • patients who have died quickly after the surgery are ”too cheap”

  8. A propensity score approach to comparing medical costs between hospital districts

  9. A propensity score approach to comparing medical costs between hospital districts

  10. A propensity score approach to comparing medical costs between hospital districts The propensity score • Rosenbaum & Rubin, Biometrika 1983 • e(x)=P(y=1|x), where y = 1 indicates that a patient with observed covariates x will be assigned to a given hospital district (y = 0, a control district) • subclassification on the propensity score will balance x, in the sense that within subclasses that are homogeneous in e(x), the distribution of x is the same for study and control districts • P(x,y|e)=P(x|e)P(y|e)

  11. A propensity score approach to comparing medical costs between hospital districts Calculating the propensity score using a logit-model • consider the probability π(x) that a patient is assigned to a given hospital district as a function of covariates x • P(y=1|x) =π(x); compare study district (y=1) vs. control (y=0) • we use the model: logit[π(x)]=log[π(x)/(1- π(x))]=α+β’x

  12. A propensity score approach to comparing medical costs between hospital districts Calculating the propensity score using the logit model… • the explanatory variables include: age, sex, days of care, type of fracture, type of surgery, did the patient come from home, days in care during the preceding 60 days, did the patient die during the follow up, days alive during the follow up and most important comorbidities classified into 13 classes • the predicted values were calculated and used as the propensity score

  13. A propensity score approach to comparing medical costs between hospital districts The costs • the patients were sorted by the propensity score within the hospital districts • the costs were smoothed over the propensity score using local linear regression with normal weights w and h as the standard deviation of the normal distribution used • the estimator:

  14. A propensity score approach to comparing medical costs between hospital districts The costs of hosp. dist. no. 1 (green) and hosp. dist no. 2 (black)

  15. A propensity score approach to comparing medical costs between hospital districts The costs of hosp. dist. no 1 (green) and hosp. dist. no 3 (black)

  16. A propensity score approach to comparing medical costs between hospital districts Calculating the propensity score using a multinomial logit model • the estimation can be done using the whole data • all of the cost curves are comparable with each other and not only pairwise as in the case of the binomial logit –model • let πj(xi) denote the probability of response j, j=1,…,J at the ith setting of values of k explanatory variables xi=(1,xi1,…,xik)’. log[πj(xi) / πJ(xi)]= βj’xi , j=1,…,J-1 , where the last category J is the baseline

  17. A propensity score approach to comparing medical costs between hospital districts Calculating the propensity score using a multinomial logit model • again the predicted values were calculated and the logit value corresponding to the probability hospital district 1 (say) is used as the propensity score • now the patients with the same (or almost the same) propensity score should have similar distribution of covariates x within every district

  18. A propensity score approach to comparing medical costs between hospital districts The costs of hpd. 1(green), 2(blue) and 3(black)

  19. A propensity score approach to comparing medical costs between hospital districts Prolonged in-hospital waiting time and costs • patients should be operated as soon as possible after the fracture, provided that their condition is sufficiently good • there is indication that a waiting time over three nights increases mortality • the patients are divided into prolonged (waiting time three nights or longer) and early (waiting time less than three nights) surgery within the hospital districts • the costs were again smoothed inside these subclasses

  20. A propensity score approach to comparing medical costs between hospital districts The cost of patients with ≥3 nights (green) waiting time and <3 nights (black) waiting time in hpd 1

  21. A propensity score approach to comparing medical costs between hospital districts The cost of patients with ≥3 nights (green) waiting time and <3 nights (black) waiting time in hpd 2

More Related