EP2440121A1 - System and method for medical treatment hypothesis testing - Google Patents
System and method for medical treatment hypothesis testingInfo
- Publication number
- EP2440121A1 EP2440121A1 EP10804855A EP10804855A EP2440121A1 EP 2440121 A1 EP2440121 A1 EP 2440121A1 EP 10804855 A EP10804855 A EP 10804855A EP 10804855 A EP10804855 A EP 10804855A EP 2440121 A1 EP2440121 A1 EP 2440121A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- treatment
- patient
- identified
- data
- identifying
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
Definitions
- the present invention relates generally to methods and systems for the statistical analysis of retrospective medical data and more specifically to methods and systems for the evaluation of the effects of medical treatments.
- the invention relates to a software component system and method for the statistical analysis of patient medical records for the purpose of evaluating the relative likelihood that patient experience identified effects when treated using identified treatments. This method of analysis produces results that decision makers in the medical field can rely on to evaluate the efficacy and risks of identified treatments in determining which treatments to approve, use, and fund.
- the system evaluates two treatments to determine the relative likelihood that patients receiving the first treatment will experience the identified effect by comparison to patients receiving the second treatment.
- the treatments being compared are actually dosage amounts of the same treatment.
- the system compares a first treatment to several different second treatments by performing several individual comparisons to produce several different ratio of rate ratios, each representing the relative likelihood that the patient will experience the identified effect as between the first treatment and another treatment.
- One embodiment of the present invention provides a computer-implemented system and method of evaluating the effects of medical treatments, the method including receiving patient record data, identifying relevant characteristics for evaluation, identifying a first treatment, identifying a second treatment; assigning a weight to each patient case, determining the relative likelihood, using the assigned weights, that an identified treatment will result in an identified effect when contrasted with a second identified treatment, and outputting this estimated relative likelihood,
- the invention in another embodiment relates to a system for evaluating the effects of medical treatments including: a network interface, a patient record database residing on a server accessed through a network, and a server for analysis, wherein the analysis server is configured to receive patient record data, identify relevant characteristics for evaluation, identify a first treatment, identify a second treatment, assign a weight to each patient case based on the likelihood that the patent would be a member of the exposed group, determine the relative likelihood, using the assigned weights, that an identified treatment will result in an identified effect when contrasted with a second identified treatment, and, output this estimated relative likelihood.
- the invention in another embodiment relates to a server for evaluating the effects of medical treatments, the server including: a processor; and memory operably coupled to the processor storing programming instructions therein, the processor being operable to execute program instructions, the program instructions including: assigning a weight to each patient case based on the likelihood that the patent would be a member of the exposed group; using the assigned weights in determining the relative likelihood that an identified treatment will result in an identified effect when contrasted with a second identified treatment.
- FIG, 1 is a schematic block diagram of a medical treatment hypothesis testing system constructed according to an embodiment of the present invention for comparing the effects of medical treatments based on retrospective observational data;
- FIG. 2 is a high-level flow diagram of the steps exercised by the data selection component of the medical treatment hypothesis testing system of FIG 1 according to an embodiment of the present invention
- FIG. 3 is a high-level flow diagram of the steps exercised by the propensity scoring component of the medical treatment hypothesis testing system of FIG 1 according to an embodiment of the present invention
- FIG. 4 is a flow diagram of the steps exercised by the Prepare Data in Table component of the propensity scoring component of FIG 3 according to an embodiment of the present invention
- FIG. 5A is an illustration of an exemplary patient case data table at the beginning of the Prepare Data in Table process of FIG. 4 according to an embodiment of the present invention
- FIG. 5B is an illustration of an exemplary patient case data table at one point in the Prepare Data in Table process of FIG. 4 according to an embodiment of the present invention
- FIG. 5C is an illustration of an exemplary patient case data table at one point in the Prepare Data in Table process of FIG. 4 according to an embodiment of the present invention
- FIG. 5D is an illustration of an exemplary patient case data table at one point in the Prepare Data in Table process of FIG. 4 according to an embodiment of the present invention
- FIG. 5E is an illustration of an exemplary patient case data table at one point in the Prepare Data in Table process of FIG. 4 according to an embodiment of the present invention
- FIG. 6 is an illustration of an exemplary table for the comparison of non-weighted patient data for patients in the treatment group and weighted patient data for patients not in the treatment group according to an embodiment of the present invention
- FIG. 7 is an illustration of an exemplary patient case data table for use in the regression modeling 118 of FIG. 1. This illustration shows the data as it is formatted during the format data for regression step 312 of FIG 3.
- embodiments of the present invention that include patient data selection, propensity scoring, inverse-probability weighting, and doubly robust estimation of the ratio of rate ratios are shown.
- Retrospective observational medical record data are selected by an operator from a patient record database based on two identified treatments and identified effects to be studied.
- Each patient who did not receive the first identified treatment is assigned a propensity score that represents the likelihood, based on the patient's individual characteristics, that the patient would have been in the group that received the first identified treatment.
- the assigned propensity scores are used to weight, each patient's data so that the weighted data of the group that did not receive the first identified treatment (i.e.
- the group that received the second identified treatment closely resembles the un- weighted data of the group that received the first treatment.
- the weighted data is used to perform a regression to estimate the relative likelihood that a patient receiving the first identified treatment, would experience an identified effect as compared to a patient receiving the second identified treatment. This estimate is called the ratio of rate ratios.
- the estimate of the ratio of rate ratios is doubly robust.
- F ⁇ G. 1 is a schematic block diagram of a medical treatment hypothesis testing system constructed according to an embodiment of the present invention for comparing the effects of medical treatments based on retrospective observational data.
- the system includes an operator terminal 102 for operator access, over a network 104 to an analysis machine 106 and a patient data machine 108.
- the patient data machine 108 provides access for an operator to a patient record database 110 and an auxiliary database 112.
- the analysis machine 106 may include a processor; and memory operably coupled to the processor storing programming instructions and other data therein, the processor being operable to execute program instructions, a network connection to permit the analysis machine to receive input from the patient data machine 108 and other sources and to output results to the operator terminal 102 and other destinations.
- the patient record database 110 stores retrospective observational patient data from patient medical records such as patient identification number, date of birth, gender, the results of medical tests, observational data recorded by healthcare providers, and information provided by patients to healthcare providers.
- the auxiliary database 112 stores information related to the operation of the system such as information about what records have been retrieved from the patient record database and when they were retrieved, system data formatting rules, and other data pertinent to the analysis of the patient record data.
- the analysis machine 106 provides access for an operator to several software components including a data selection component 114, a propensity scoring component 116, a regression modeling component 118 and a data output component 120. These software components may take the form of computer instructions stored in computer memory and executed by a computer processor.
- the data selection component 114 presents the operator with an interface for the selection of relevant patient data and attributes from the patient record database 110, and retrieves and formats the selected data for use in the propensity scoring component 116.
- the propensity scoring component 116 determines and assigns a propensity score to each patient record that represents that patient's likelihood of being in the group of patients receiving the identified treatment ("in the treatment group"). The propensity scoring component 116 further applies the propensity score to the patient data to weight the data of the patient records such that the weighted data for the group of patients not in the treatment group closely resembles the non-weighted data for the group of patients in the treatment group.
- the regression modeling component 118 provides an interface for the operator to build and test a model for estimating the relative likelihood that a patient receiving the first identified treatment would experience an identified effect as compared to a patient receiving the second identified treatment (i.e. a patient not in the treatment group).
- the regression modeling component 118 receives weighted data weighted by the propensity scoring component 116.
- the data output component 120 provides an interface to allow the operator to select the format and style for presenting analysis machine results.
- the data output component 120 includes a tool for selecting and formatting data produced by the analysis machine.
- the tools in the data output component 120 allow the operator to select and manipulate various visualization tools such as charts and graphs to assist interpretation and understanding of analysis machine results.
- FIG. 2 is a high-level flow diagram of the steps exercised by the data selection component 114 of the medical treatment hypothesis testing system of FIG 1 according to an embodiment of the present invention.
- the data selection component 114 initially establishes a connection to the patient record database 202 to provide access to the patient data for the operator.
- the component utilizes an ODBC driver to connect the data selection component 114 to the patient record database 110 on the patient data machine 108 across the network 104.
- the component then provides a data selection interface to the operator 204 to allow the operator to select the appropriate patient records for the subsequent analysis on the analysis machine.
- the data selection interface is presented to the operator as an HTML webpage viewed through a web browser application.
- the system receives this selection 206 and retrieves the selected records 208 from the patient record database 110.
- the system then formats the data 210 for use in the analysis machine by the propensity scoring component and stores the formatted data 212 in memory on the analysis machine 106.
- FIG. 3 is a high-level flow diagram of the steps exercised by the propensity scoring component 116 of the medical treatment hypothesis testing system of FIG 1 according to an embodiment of the present invention.
- the propensity scoring component 116 performs an initial data preparation step 302 described in greater detail with respect to FIG, 4.
- the analysis machine prepares the patient records in a table and identifies the initial "seeding" propensity score and residual associated with each patient record, creates columns for and evaluates each indicator function and the products of indicator functions for each patient characteristic described in the selected data.
- the analysis machine identifies the largest absolute correlation 304 between any indicator function column or product of indicator functions column and the residual column.
- f,- represents the 0/1 indicator of patient i receiving the first identified treatment, /?, • is the estimated probability that patient i received the first identified treatment, I j is the/ h indicator function, and n is the number of patients.
- fo ⁇ nula is employed: ⁇ - Sum ⁇ ⁇ t t - Pi)(Iji ⁇ mean(/ / ))/((n - l)sd(t - p)sd(I / )).
- the analysis machine determines a weight to apply to each of the non-treated patient records 308.
- the weight is calculated from the propensity score using the following formula: w,-
- the analysis machine compares the aggregate weighted data for patients not in the treatment group against the aggregate data for patients in the treatment group 309 to determine whether the two data sets are sufficiently similar (“optimally balanced") 310. ⁇ f the data sets are not optimally balanced, the analysis machine returns to the data table, that contains newly assigned propensity scores, assigned in the adjustment process 306 and the newly calculated residual (adjusted by the change of the propensity score).
- the analysis machine then repeats the process of identifying the largest absolute correlation 304 with the new residual, adjusting the patient propensity scores 306, determining new patient data weights based on the newer propensity scores 308 and comparing the weighted non-treatment data with the non-weighted treatment data 309 to determine whether the data sets are now optimally balances 310. This process repeats until the data sets are sufficiently similar, at which point the analysis machine formats the data for use in the regression step 312.
- FIG. 4 is a flow diagram of the steps exercised by the Prepare Data in Table component of the propensity scoring component of FIG. 3 according to an embodiment of the present invention.
- the data preparation steps depicted in FIG. 4 interpret and arrange the selected patient data in a manner such that it can be interpreted by the subsequent step in the propensity scoring process involving correlation 304.
- the data preparation described in the foregoing steps detailed in FIG. 4 is also illustrated in exemplary tables in FIG. 5A - FIG. 5E.
- FIG. 4 shows the system initially retrieves the patient case data 402 that was stored within the data selection component process 114 also depicted in detail in FIG. 2.
- An exemplary table containing patient case data for five example patients is illustrated in FIG. 5A.
- each patient record at the initial stage may contain at least a patient ID number, a treatment indicator for identifying whether the patient was a member of the treatment group and data associated with categorical features such as race as well as numeric measurements such as blood pressure.
- the next step in the process is the assignment of an initial "seeding" propensity score to each patient 404.
- this seeding propensity score is calculated by the analysis machine dividing the number of patients within the treatment group by the total number of patient records.
- the next step in the data preparation process is the calculation of the residual for each patient record. In one embodiment of the present invention, the residual is calculated by subtracting the propensity score from the treatment indicator for each patient.
- Exemplary tables containing propensity score and residual data for each of the five example patients are illustrated in FIG. 5B and FIG. 5C respectively.
- the analysis machine then defines and evaluates indicator functions for each patient characteristic described in the data.
- the definition of said indicator functions, and creation of associated columns, is illustrated in the exemplary table of 5D that shows a new column for each condition that must be defined to allow the analysis machine to describe in boolean terms both categorical features and numeric measurements.
- the evaluation of all indicator functions is avoided by the use of published algorithmic shortcuts, in this case only selected indicator functions are defined and evaluated.
- the indicator functions are evaluated by determining whether the condition contained in the indicator function is true or false with respect to each patient record, with a 1 or 0 placed in the column to represent the same.
- the next step in the process of FIG 4 is the definition and evaluation of the product of each indicator function with each other indicator function 410.
- the system creates a new column for each condition and evaluates each condition placing a 0 or 1 in the relevant column to indicate whether the patient identified in that row meets the condition identified in the indicator function, in this case the product of one or more indicator functions, at the top of each column.
- the columns created by this step and the evaluation of patient records is illustrated in FIG 5E.
- FIG. 6 is an illustration of an exemplary table for the comparison of patient data for patients who received the treatment and weighted patient data for patients who did not receive the treatment according to an embodiment of the present invention.
- FIG. 6. illustrates the data considered in the comparison step 309 of FIG. 3 in determining whether the data has reached an optimal balance 310.
- the data in the exemplary table of FIG. 6 shows a sufficiently balanced data set after weighting to prepare the data for use in regression 312.
- FIG. 7 is an illustration of an exemplary patient case data table for use in the regression modeling 118 of FIG. 1. This illustration shows the data as it is formatted during the format data for regression step 312 of FIG 3. In this example the original data from FIG. 5 A is supplemented with data on the observed identified effects.
- the table of FIG. 7 shows two records for each patient, one for a "pre-treatrnent period" and one for a "post-treatment period.”
- the "period" column of FIG. 7 contains a 1 or 0 to indicate whether the record describes the "post-treatment” or "pre-treatment” period respectively.
- the table of FIG 7. also includes a column/indicator for propensity score weight, "W”, computed by the system 308 as depicted in FIG 3 and described above.
- the relative likelihood of a patient receiving the first treatment experiencing the identified effect when compared with a patient receiving the second treatment is calculated by estimating the ratio of rate ratios.
- the ratio of rate ratios (“RRR") compares the rates of the identified event in the pre-treatment period to the post-treatment period for the treatment (first treatment) and non-treatment (second treatment) groups.
- the formula for calculating the RRR is : (rate(post,treatment)/rate(pr ⁇ ,treatment))/(rate(post,non-treatment)/rate(pre,non- treatment)).
- the analysis machine outputs this estimated RRR to the operator terminal or another location where it may be used to evaluate the relative efficacy and risks of the first and second identified treatments. Decision makers in the medical field may utilize the estimated RRR output from the analysis machine to determine whether to approve or recommend particular treatments for patients and healthcare organizations.
- the analysis machine After calculation, the analysis machine outputs the doubly robust adjusted ratio of rate ratios to the operator terminal for review by the operator, or to another location where it may be used to evaluate the relative efficacy and risks of the first and second identified treatments. Decision makers in the medical field may utilize this doubly robust adjusted ratio of rate ratios output from the analysis machine to determine whether to approve or recommend particular treatments for patients and healthcare organizations.
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US18675909P | 2009-06-12 | 2009-06-12 | |
PCT/US2010/038561 WO2011014308A1 (en) | 2009-06-12 | 2010-06-14 | System and method for medical treatment hypothesis testing |
Publications (2)
Publication Number | Publication Date |
---|---|
EP2440121A1 true EP2440121A1 (en) | 2012-04-18 |
EP2440121A4 EP2440121A4 (en) | 2014-12-10 |
Family
ID=45812051
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP10804855.4A Ceased EP2440121A4 (en) | 2009-06-12 | 2010-06-14 | System and method for medical treatment hypothesis testing |
Country Status (1)
Country | Link |
---|---|
EP (1) | EP2440121A4 (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002088901A2 (en) * | 2001-05-02 | 2002-11-07 | Qed Solutions, Inc. | A method for analyzing drug adverse effects employing multivariate statistical analysis |
EP1717722A2 (en) * | 2005-04-25 | 2006-11-02 | Ingenix Inc. | System and method for early identification of safety concerns of new drugs |
US20080082582A1 (en) * | 2006-09-29 | 2008-04-03 | Searete Llc, A Limited Liability Corporation Of The State Of Delaware | Computational systems for biomedical data |
-
2010
- 2010-06-14 EP EP10804855.4A patent/EP2440121A4/en not_active Ceased
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002088901A2 (en) * | 2001-05-02 | 2002-11-07 | Qed Solutions, Inc. | A method for analyzing drug adverse effects employing multivariate statistical analysis |
EP1717722A2 (en) * | 2005-04-25 | 2006-11-02 | Ingenix Inc. | System and method for early identification of safety concerns of new drugs |
US20080082582A1 (en) * | 2006-09-29 | 2008-04-03 | Searete Llc, A Limited Liability Corporation Of The State Of Delaware | Computational systems for biomedical data |
Non-Patent Citations (1)
Title |
---|
See also references of WO2011014308A1 * |
Also Published As
Publication number | Publication date |
---|---|
EP2440121A4 (en) | 2014-12-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8538906B2 (en) | System and method for medical treatment hypothesis testing using relative likelihood of treatment results | |
Bhattacharya et al. | Estimating probit models with self‐selected treatments | |
US10216901B2 (en) | Auditing the coding and abstracting of documents | |
US20070179769A1 (en) | Medical risk stratifying method and system | |
US8793144B2 (en) | Treatment effect prediction system, a treatment effect prediction method, and a computer program product thereof | |
CA2470733C (en) | Generation of continuous mathematical model for common features of a subject group | |
KR102131973B1 (en) | Method and System for personalized healthcare | |
CN112331285B (en) | Case grouping method, case grouping device, electronic equipment and storage medium | |
Higgins et al. | Meta-regression | |
CN110704583A (en) | Medical record analysis method and device, server and storage medium | |
US20140019090A1 (en) | Systems, methods, and logic for generating statistical research information | |
EP3718116B1 (en) | Apparatus for patient data availability analysis | |
CN111968740B (en) | Diagnostic label recommendation method and device, storage medium and electronic equipment | |
CN112633321A (en) | Artificial intelligence recommendation system and method | |
de Castro et al. | Hypothesis testing in an errors‐in‐variables model with heteroscedastic measurement errors | |
EP2440121A1 (en) | System and method for medical treatment hypothesis testing | |
US20230260659A1 (en) | Method for a predictive prognosis of the onset of a cardiovascular disease | |
EP4165646A1 (en) | Methods and systems for searching an ecg database | |
EP3506268A1 (en) | Apparatus for patient data availability analysis | |
Johnson | Analysis of a Medical Center's Cardiac Risk Screening Protocol Using Propensity Score Matching | |
Thompson et al. | Group-level analyses involving scores linked from legacy scales to PROMIS scales: A novel alternative using imputation. | |
CN117373665A (en) | Digital health management method based on big data and related device | |
CN117546250A (en) | Systems and methods for estimating treatment efficacy using covariate adjustment stratification and pseudo-value regression in randomized trials | |
CN116913447A (en) | Method and device for processing dry weight information, electronic equipment and medium | |
CN112837814A (en) | Data display method, device, equipment and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
17P | Request for examination filed |
Effective date: 20120109 |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO SE SI SK SM TR |
|
DAX | Request for extension of the european patent (deleted) | ||
A4 | Supplementary search report drawn up and despatched |
Effective date: 20141110 |
|
RIC1 | Information provided on ipc code assigned before grant |
Ipc: G06F 19/00 20110101AFI20141104BHEP |
|
17Q | First examination report despatched |
Effective date: 20170915 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R003 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE APPLICATION HAS BEEN REFUSED |
|
18R | Application refused |
Effective date: 20190116 |