For each measure, health centers have the option of reporting on their entire patient population as a universe. To report on the universe, the data source such as an Electronic Health Record must include all medical patients from all service delivery sites and grant funded programs (e.g., CHC, HCH, MHC, PH) in the defined universe.  In addition, the data source must cover the period of time to be reviewed (e.g., three years for pap tests, etc.) and include information to assess compliance with the clinical measure as well as to evaluate exclusions.  Reporting on the universe is more accurate (i.e., it reports on 100% of patients) and easier (i.e., queries are automated).   Although optimal, there is no requirement that health centers report the universe.  Indeed, BPHC has no preference for reporting the universe or a sample.

If the health center cannot report on the universe (or chooses not to), a random sample will be used to report.  Note that the health center can report on the universe for some measures while using a sample to report others.  It is not necessary that all measures be reported using the same method. 

The following measures can be reported using a sample:

• Table 6B Childhood Immunization Rate – percent of patients 2 years of age with up-to-date immunizations.
• Table 6B PAP Test Rate – percent of female patients aged 21 – 64 who have had a PAP test during the measurement year or during the previous two calendar years.
• Table 7 Controlled Diabetes – Percent of diabetic patients with HbA1c levels
• Table 7 Controlled Hypertension – percent of patients 18 years of age and older with hypertension whose latest blood pressure was less than 140/90.

Table 6B and 7 prenatal indicators cannot be reported using a sample.

RANDOM SAMPLE

A random sample is defined as a part of a universe where each member of the universe has the exact same chance of being selected as every other member of the universe.

Thus, a true random sample will generate outcomes which are similar to outcomes reported for the universe of patients because the sample is "representative" of the universe.

STEP BY STEP PROCESS FOR REPORTING CLINICAL MEASURES

For each measure, perform each of the following steps.

STEP 1: Identify the patient population to be sampled (the universe):

Define the universe for each condition.

• Including all active medical patients
  
• Including all sites in the scope of project
  
• Including contracted medical services

Identify the number of patients who fit, or who initially appear to fit, the criteria for that measure.  Create a list and number each member of the patient population in the universe. 

STEP 2:  Determine the sample size for manual chart review:

The number of charts selected for manual chart review will be the lesser of 70 charts or all patients who meet the criteria.

STEP 3:  Select the random sample

Using one of the two recommended sampling methodologies, identify the sample of 70 charts (assuming the universe is greater than 70, otherwise report on all patients).

STEP 4:  Review the sample of records to determine compliance with the clinical measure.

For each measure, review available data sources to identify any automated sources to simplify data collection.  Since these data sources will be augmented by the paper record, they do not need to include all patients from all service sites and programs.  Examples of data sources include:

• Electronic health record
  
• PECs database
  
• State immunization registries for vaccine histories
  
• Logs
  
• Practice management system

For each patient in the sample, determine whether sufficient information is available in available data source(s) to assess compliance.  If information is not available, pull the paper record to retrieve required information. 

STEP 5:  Replacing patients that should be excluded from the sample.

If a patient is selected that should be excluded from the sample, the patient will be replaced with a substitute.  Use the replacement methodology described for the sampling methodology selected.  Exclusions are as follows:

• Childhood immunizations – none
  
• Pap tests – women who have had a hysterectomy
  
• Controlled hypertension – none
  
• Controlled diabetes -  patients with a diagnosis of polycystic ovaries that do not have two face-to-face encounters with the diagnosis of diabetes, in any setting, during the measurement year or year prior to the measurement year; gestational diabetes (ICD-9-CM Code 648.8); or steroid-induced diabetes (ICD-9-CM Code 962.0, 251.8) during the measurement year.

Using this method, the final sample size to be reported on the UDS will be the lesser of 70 charts or all patients who meet the criteria for each measure but never more than 70.

METHODOLOGY FOR OBTAINING A RANDOM SAMPLE

Two methods are recommended for generating a random sample and replacements for excluded patients:

• Work with a list of random numbers generated for your total patient population.
  
• Select a random starting point and use a calculated interval to find each next member of the sample.

Either method can be used to create a "replacement list" used to replace patients who are excluded.

Option #1: Random Number List

A list of random numbers can be created at the web site: http://www.randomizer.org/form.htm

The web site requires no password or subscription to access.  To obtain a list of random numbers, complete the questions as documented below.   Complete the "Number Range" by entering the maximum number of patients in the universe for the particular measure under consideration as "n".  For example, if there are 700 children who turn two in the reporting year in the universe, enter 700 as the maximum range.

Then press the button "Randomize Now!"  A list of randomly generated numbers will be created.  These numbers correspond with the numbered list of patients in the universe prepared in Step 1, above.

Identifying a replacement:

To create a "sample" of patients to substitute for patients who should be excluded from the sample, follow the instructions for creating a list of random numbers for a replacement sample.  Rather than selecting 70 numbers for the set, select a small sample of 5 charts.  If a patient should be excluded from the original random sample of 70, replace that patient with one of the patients from the replacement sample.  In this manner, more than 70 patients may be evaluated for compliance for a particular measure but the final sample will include 70 patients who meet all the selection criteria.   

Option #2: Interval

A second method uses the same numbered list of patients in the universe created in Step 1, above.  To generate the sample:

1. Calculate sampling interval by dividing number of patients in the universe by 70:

Sample Interval Size (S1) =  Population size (number in universe)/ Sample size (70)

2. Randomly pick a patient from the first sampling interval. For example, if the sampling interval is 10, the first sampling interval includes patients no.1 through no.10.  Randomly select one patient from this interval.

3. That will be your first record sequence number

4. Then, select every nth patient based on the sampling interval until you reach the desired sample size. In our example, if the first patient selected is number 8, and the sampling interval is 10, then the remaining patients to be selected are no.18, 28, 38, etc.

first sequence # + SI = second #

5. Continue through list until all 70 have been identified

Identifying a replacement:

If a selected patient should be excluded from the sample, return to the original list and substitute the excluded patient by the next patient on the list.  If that patient should be excluded select the next patient on the list until an eligible patient is selected.  Resume selection using the next chart you had pre-selected for the sample.  (If you run out of patients, continue your count back at the beginning of the universe).   In this manner, more than 70 patients may be evaluated for compliance for a particular measure but the final sample will include 70 patients who meet all the selection criteria.