Appendix I:
Measure Sampling
Introduction
Sampling is a process of selecting a representative part of the population of interest in
order to estimate the organizations performance, without collecting data for its entire
population. Using a statistically valid sample, an organization can measure its
performance in an effective and efficient manner.
To obtain statistically valid sample data, the sample size should be carefully determined
and the sample cases should be randomly selected such that individual cases in the
population have an equal chance of being selected. Only when the sample data truly
represent the whole population can the sample-based performance measure data be
meaningful and useful. Guidelines for effective sampling procedures follow.
Sampling may be used for all measures in the Library of Measures except for I-HBIPS-
2, I-HBIPS-3, I-NCS-2, I-NSC-4 and I-NSC-5. These measures cannot be sampled
because they are event-based measures.
Sampling Availability
If an organization decides to sample a measures or measure sets initial eligible
population, sampling should be applied to all monthly discharge medical records
identified as part of the inpatient initial eligible population. Initial eligible population
criteria are described in each measure set chapter of the specification manual.
The initial eligible population should be identified by using available databases or other
information repositories that contain monthly patient discharge information, International
Classification of Diseases (ICD) diagnosis/procedure codes or patient diagnoses or
procedures, and other necessary administrative data (e.g., patient age).
A hospital has 57 discharges (N) in a month for a given measure initial patient
population. All 57 cases (100%) should be reviewed. Sampling is not
appropriate in this example.
A hospital has 128 discharges (N) in a month for a given measure initial patient
population. The sample size for this month would be 58 cases (n).
A hospital has 512 discharges (N) in a month for a given measure initial patient
population. The sample size for this month would be 58 cases (n).
A hospital has 905 discharges (N) in a month for a given measure initial patient
population. The sample size for this month would be 58 cases (n) (the maximum
required number of cases for the monthly sample size).
such a way that a sample size of n is obtained, where k < N/n. The first sample record
(i.e., the starting point) must be randomly selected before taking every kth record. This
is a two-step process as follows:
For a measure or measure set with an initial patient population size of 360 discharges
per month (N), the sample size would be 58 (n) according to the table provided in this
section. To select a random sample of 58 cases you would implement the following
process:
1. Determine the initial patient population size N (i.e., the total number of
discharges associated with the selected measure) for the month.
2. Determine the sample size n using the above table.
3. Divide the population size N by the suggested sample size n and the quotient
is k (i.e., the resulting integer is the sampling interval k).
Example: How to get sample interval number
a. The sampling interval k = 360/58 = 6 sampling interval (k)
b. Thus, every 6th (k) patient record will be selected from the measure
population until 58 cases have been selected.