Data quality considerations

The percentage of Q1 questionnaires linked to HES is unlikely to reach 100% completeness due to the voluntary nature of PROMs for patients. There are several additional factors that will also affect the level of matching between HES and PROMs, both for HES episodes that do not link to PROMs and vice versa. Some of these reasons are listed below.

Completion of PROMs is voluntary: patients are not obliged to take part in the PROMs programme at either Q1 or Q2 stage.

Provider mapping issues: the Q1 may be administered to a patient at one hospital and the operation performed at another. The matching algorithm can take many of these provider-to-provider relationships into account where they are known but there will always be relationships between providers that are not recognised.

Lag or ‘lead in’ time before the operation: The Q1 may have been completed more than 18 weeks before the operation.

Poorly completed patient details on the questionnaire: patients are asked to complete their personal details on the Q1 and these details are sent to the Personal Demographics Service (PDS) in order to return the NHS number for matching purposes, a process called patient tracing. In some cases, the quality of the details supplied by the patient are not of sufficient quality for successful tracing. A common example of this is where the patient enters the current date when asked for their date of birth.

Un-coded or poorly coded episodes: only episodes with codes in the list (see Appendix 4 and 5 in the accompanying excel file) are included in the pool of operations eligible for matching. Often the clinical coding in new episodes are either poorly completed or missing altogether which will cause the episode to become ineligible for inclusion. 

Six months after the procedure, the supplier sends out the follow-up post-operative questionnaire. Once returned it is scanned and linked with the pre-operative and HES data. All post-operative questionnaires are linked to their pre-operative questionnaire through the PROMs serial number, which is the same for both questionnaires.

Where a patient has declined to answer a question, a default value is inserted to represent a null value. Missing responses to any of the elements within a scoring system mean that the total cannot be calculated. An exception to this is the Oxford orthopaedic scores. PROMs programme imputes missing values as the average of present responses provided that no more than two responses are missing. Where more than two responses are missing, the overall score is not calculated.

Where a patient has indicated more than one value within a single multiple-choice question or EQ-5D^TM VAS the most sever answer will be chosen. For example, if a patient has answered 3 (very mild) and 1 (moderate) to question one of the knee replacement specific questions: ‘During the past 4 weeks, how would you describe the pain you usually had from your knee?’, then 1 would be chosen in preference to 3.