How to investigate hospital mortality statistics Mohammed A

Outline • Challenging work • Signal from Noise • Scientific process • Hypothesis generation

Hospital Mortality Statistic • Hospital Standardised Mortality Ratio • Observed Number of Deaths (O)

Case-study • Shropshire and Telford NHS Trust • Dr Foster HSMR • 99 (2008/9)

Predictions (theory) • HSMR increase due mainly to increase in Observed deaths then: •

Investigation method (1) Separately consider the numerator (observed deaths) and denominator (expected deaths) of

Data • We used Sa. TH hospital admissions data • (n= 74, 860) from

Table 1. HSMRs for Sa. TH and its two constituent hospitals over three years.

The Less Severe Admissions Hypothesis 10

Visualising data (using simple run charts) “It provides a front line of attack, revealing

Reflections • Asking why as HSMR is high is very easy • Distinction between

Slides: 12

Download presentation

How to investigate hospital mortality statistics Mohammed A Mohammed 1

Outline • Challenging work • Signal from Noise • Scientific process • Hypothesis generation and hypothesis testing • Pyramid Model of Investigation • Tools • Visualisation using simple run charts • Test for interactions (BMJ 2009) • Case-study 2

Hospital Mortality Statistic • Hospital Standardised Mortality Ratio • Observed Number of Deaths (O) • Expected (from a statistical model) number of deaths (E) • HSMR = O/E *100 • An elevated HSMR could come from • Increase in O • Decrease in E (due to poor care, …) (due to less sick patients, …)

Case-study • Shropshire and Telford NHS Trust • Dr Foster HSMR • 99 (2008/9) increased to 118 (2009/10 – “outlier”) • Two hospitals • RSH and PRT • Why? Mohammed MA, Stevens AJ (2013) A Simple Insightful Approach to Investigating a Hospital Standardised Mortality Ratio: An Illustrative Case. Study. PLo. S ONE 8(3): e 57845. doi: 10. 1371/journal. pone. 0057845 http: //www. plosone. org/article/info: doi/10. 1371/journal. pone. 0057845 4

Predictions (theory) • HSMR increase due mainly to increase in Observed deaths then: • (1) A deterioration in quality of care (not a desktop exercise) • (2) an increase in patient severity not reflected in the expected mortality. • HSMR increase due mainly to fall in Expected deaths then: • (1) A genuine decrease in patient severity accompanied by a deterioration in quality of care (not a desktop exercise) • (2) that the case-mix adjustment model is underestimating the risk of admissions (eg perhaps because of inadequate clinical coding). 5

Investigation method (1) Separately consider the numerator (observed deaths) and denominator (expected deaths) of the HSMR • because this shapes the subsequent investigation process. (2) Visualise the data using simple run charts, • because visualisation is critical to data analysis • “It provides a front line of attack, revealing intricate structure in data that cannot be absorbed in any other way. ” Cleveland WS (1993) Visualizing Data. (3) Make comparisons with other hospitals over the same time period • enabling a form of controlled comparison. 6

Data • We used Sa. TH hospital admissions data • (n= 74, 860) from the Dr Foster Real Time Monitoring computer system • Three year period from April 2007 to March 2010 • 36 months. 7

Table 1. HSMRs for Sa. TH and its two constituent hospitals over three years. 8

Numerator Denominator Plots 9

The Less Severe Admissions Hypothesis 10

Visualising data (using simple run charts) “It provides a front line of attack, revealing intricate structure in data that cannot be absorbed in any other way. ” Cleveland WS (1993) Visualizing Data.

Reflections • Asking why as HSMR is high is very easy • Distinction between excess deaths and avoidable mortality not made explicit • Answer is not straight forward • Desktop exercise • Signal from noise • Often large number of analyses • Credible explanations • Often not conclusive but moderate/high degree of belief • Question of avoidable deaths remains unanswered • Case-note review • Provides approximate answer to the “killer” question • …. . 12