Development of a predictive model algorithm to identify patients with hypophosphatasia, using Optimum Patient Care Record Database in United Kingdom

Study identification

• EU PAS number:
  EUPAS44118
• Study ID:
  47750
• Official title and acronym:
  Development of a predictive model algorithm to identify patients with hypophosphatasia, using Optimum Patient Care Record Database in United Kingdom
• DARWIN EU® study:
  No
• Study countries:
  United Kingdom
• Study description:
  The study is a retrospective observational case-control study accessing de-identified primary healthcare records from patients enrolled in the Optimum Patient Care Record Database in the UK. The study observation period will start at 1st January 2000 and end on 31st March 2021. HPP cases will be identified based on Read or Systematized Nomenclature of Medicine Clinical Terms (SNOMED-CT) codes, with index date defined as date of first HPP diagnosis during the study period. Controls will be a random selection of non-HPP patients matched by year of birth/age, gender, date of earliest record of index case and being alive at index date. Controls will be collected with a target ratio of 1 case to 20,000 controls. The pooled cases and controls will be randomly allocated to a (i) training (75%), or (ii) validating dataset (25%). For patients’ electronic health records respectively in the training and validating datasets, predictor variables will include all available data items as Read or SNOMED-CT codes recorded any time prior to index date. A machine learning prediction model (the “scoring algorithm”) will be developed in the training dataset and will then be tested in the validating set using statistical measures of accuracy, discrimination, and calibration. The validation step will involve estimating the predicted probability of HPP diagnosis for each control patient and rank-ordering of patients according to their predicted probabilities (“score”). As a next step, at least two clinical experts in HPP will perform a chart review of the top 10% ranked patients and score patients as ‘highly likely HPP’, ‘likely HPP’, ‘unlikely HPP’, ‘highly unlikely HPP’, ‘not HPP’, or ‘unable to assess’. Based on this clinical assessment, a threshold for possible or likely HPP will be defined, and the scoring algorithm will be determined.
• Study status:
  Finalised

Research institutions and networks

Institutions

OPEN Health

Study timelines

Date when funding contract was signed

Actual: 13/02/2021

Study start date

Planned: 17/12/2021

Actual: 21/12/2021

Data analysis start date

Planned: 17/12/2021

Actual: 04/01/2022

Date of interim report, if expected

Planned: 21/12/2021

Actual: 27/01/2022

Date of final study report

Planned: 30/12/2022

Actual: 22/07/2022

Sources of funding

Pharmaceutical company and other private sector 

More details on funding

Alexion Pharmaceuticals

Regulatory

• Was the study required by a regulatory body?:
  No
• Is the study required by a Risk Management Plan (RMP)?:
  Not applicable