Health Risk Assessment And Stratification Health And Social Care Essay

In the era of introduction of programs that evaluate the risk involved in managing of disease by analyzing the risks involved through the thorough examination of clinical data collected by the help of general practitioners, pharmacies and major hospital records

As such programs have scaled to a lager extent with still a lot of scope, it has become critically essential to justify their returns. It is no longer sufficient to defend a program based on an illustrated ROI. Insurers are spectacle about which members are being identified and how, what interventions can be applied to them most effectively, and which mechanism leads to genuine behaviour change and savings [1].

This points to the fact that the insurers and program developers are just becoming smarter about the economic pay back and optimization of the disease management programs hence resulting in the sudden demand and increase in data procurement These data requirements will only increase in the future for optimization of disease management efforts. This will require intensive risk profiling, predictive modelling and stratification on the part of all who are involved in program design and execution [2].

WHAT IS RISK STRATIFICATION?

Exact definitions of risk stratification in theoretical terms are given below,

“The constellation of activities (e.g. Lab and clinical testing) used to determine a person’s risk for suffering a particular condition and need for preventive intervention.”(McGraw-Hill Concise Dictionary of Modern Medicine)

OR

“The method of delimiting sub-populations within a cohort which have different risks of a particular outcome, based upon severity of illness and co-morbidity.”(Society for Cardiothoracic Surgery in Great Britain and Ireland)

Let us acquaint ourselves with the basic terms related to Risk Stratification.

RISK:

The term risk in preoperative context include the chances of an adverse medical outcome like persistence, recurrence of the medical condition finally leading to ultimate reduction in the survival rate of the patient.

RISK ASSESMENT:

The evaluation of the clinical data collected at hospitals, GP’S, and pharmacies can be used to derive the probability of a medical risk occurring. The evaluation of this information made available through various procurement techniques. This assessment may help in identifying the level of risk, the treatment to be administered and chances of survival of the particular member.

Risk assessment helps determine the probable patients that may require immediate or future surgeries (i.e. operable patients), who would need multimodality therapy and who may require treatment under the practitioners watchful eye.

The stratification of patients divides the probable risks as low risk score which lies between 21 to 100% ,moderately relative risk that lies between 6 to 20 % and high risk between 0.5 to 5% score anything lying below a probable 0.5% lies under the high risk radar leading to a surety that the patient would require future medical care.

Figure 1 Risk Score Model

NEED OF RISK STRATIFICATION

Risk stratification provides a thorough examination of the risk of future hospital admission varying across a population where the health and social care can be intervened to patients.

Who may require it the most

Who may benefit from it the most

Hence encouraging and providing proactive healthcare to the emergencies and also support delivery of efficient service.

HOW DOES RISK STRATIFYING WORK?

Three approaches are commonly used for risk stratification:

Accurate stratification of the risk involved with the patient is a key component in health care assessment of the procedural outcomes .An increasing no of health care organizations are getting dependent on the health care assessment through such programs as a means for assistance in making patient risk-stratification decisions. The only difficulty persist that the process of outcome model development is both time consuming and difficult due to the preliminary stage.

Many techniques can be used for medical assessment of risk like modelling techniques (logistic regression, artificial neural network (ANN), and Bayesian) to rapidly develop models for risk stratifying patients. The only difference is their method of analysis. The problem pertains that none of the technique give accurate results or are or hundred percent dependable.

Threshold Analysis where a set of criteria are defined describe ‘high risk’ patients alone.

Thresholds are target based research technique based on evidence that provides numerical targets for healthy development. Targets are derived by careful evaluation of the given literature on account of the case studies agree on a particular phenomenon. Various different sub populations like

Children who are vulnerable in health conditions.

Heart patients who suffer the risk of sudden casualty and medical attention.

Diabetic patients.

Cancer patients so on.

Various such evaluations are being conducted by evaluation of patient data for all above sub populations.

Associations are conditions should be fulfilled, through a careful evaluation of the existing research literature. A committee may be formed to agree on the nature and direction of the conditions, the particular phenomena may not lend it well to the numerical thresholds precisely.

Clinical knowledge where practitioners based on their knowledge, experience and training identifies individuals who may in future prove to be ‘high risk’; and their current status. Patients keep visiting hospital and their general practitioners for pre and post operative care have their records maintained with health professional which may in the future be evaluated and help in the finding out the hospitalization risk and mortality risk of patients.

Predictive modelling the historic data made available through varying sources is used to evaluate and create an association between the patient’s current health condition and the likelihood o the patient to become a high risk in the future.

PREDICTIVE RISK STRATIFICATION (PRISM)

Predictive modelling

Is a process used in predictive analytics to create a statistical model of future behaviour? Predictive analytics is the area of data mining concerned with forecasting probabilities and trends.

[Definition from http://searchdatamanagement.techtarget.com/definition/predictive-modeling]

Predictors

A predictive model consist number of predictors that are variable factors which may tend to influence future behaviour or results. For example the age of a heart patient plays a potent role in the outcome of analysis.

In predictive modelling, data is accumulated for a relevant set of predictors, a statistical model is formulated, and on the basis of the available data predications are validated. The model may be based on a simple liner equation or may be formulated using a highly complicated neural network lattice.

Risk Stratification Tools

Risk stratification models can assist clinicians in making decisions on the subject of the need for additional testing once a preliminary clinical estimate has been performed. The American Society of Anaesthesiologists (ASA) categorization of Physical Status was the first clinical manifestation developed to forecast risk. Introduced in 1941, it was remodelled to its current form in 1962 [4]. Patients are categorized into one of 5 major classes based upon the presence and manifestations of affiliated medical disorders and whether emergency surgery is required. The utility of this index is limited by intra viewer inconsistency in rating and variations in the predictive power for postoperative hitches.

Demand of Predictive Modelling in Health Care:

A process when applied to available data identifies a person having high medical need and are “at risk” for the medical attention.

The concept is in demand for several reasons:

Most plans are saturated in data and want to use them to increase the efficiency and effective application of medical care.

In this time of severe medical inflation, it’s understandable that we focus on very high cost persons.

Case management and disease management (DM) programs are everywhere. Predictive Modelling could increase the reliability for returns of these programs.

Vendors and consultants have created a steering demand for Predictive Modelling Which is highly dependent on the marketing and scientific back grounds.

Predictive Modelling for Risk Measurement

Predictive modelling is part of a larger risk assessment and adjustment scale.

Risk factors, outcome measures, and estimation period to be related.

Predictive Modelling requires a more complex statistical engines Prediction/forecasting in medicine and healthcare is not new. Today the new reason for predictive modelling introduction is high risk case identification.

Figure 2 Stages of Prevention

Disease Stage, Prevention, and The Care Management Process are correlated in a way where at an initial stage the patient visits a GP based on early symptoms or distress at this stage the practitioner accesses the patients treatment needs or suggest certain tests t the patient. At the next stage where after evaluation of the disease the future predication is made on the need of hospitalization and hence the population is sorted into the sects required. On encountering a major disease where reoccurrence of disease or mortality is a condition a set of principles for disease management in performed by the practitioners if hospitalization is required an operative need arises case management under hospital administration is the final stage which may lead to either complete treatment of the high risk patient or mortality, complication or reoccurrence.

THUS THE NEED OF PREDICTIVE MODELLING ARISES WHEN

There is a need to identify persons for admission in intensive case-management program.

To more effectively target disease management programs and focus on providing medical attention to the persons who are in need of attention.

To provide properly calculated risk information useful in making financial decisions and budget management.

To provide information to clinicians that may prove useful for quality improvement of patient standards.

To identify the need for educational campaigns and camps for better clinical outreach programs.

STEPS IN IMPLEMENTING OF PREDICATIVE MODEL

Data Requirement

Various types of factual data like administrative records, increase and disease in records or encounter of rare or recurrent cases. The various means of data collection is through GP records, in house and out house patient records, emergency and accident case histories and so on. This phase may also be called initialization or pre processing stage.

Data warehousing or creation of a repository

The data collected over a fixed period of time of a similar format represents a warehouse. This data is collected through reliable sources. This warehouse data is subject to analysis as a processed input this marks the beginning of the stratification process.

Statistical analysis

The most important phase is to create a statistical engine with regressive research, setting conditions and associations there which the outcome is dependent. This marks the start of stratification of risked based on a scaring system that is predetermined by various conditional studies before product release.

Reporting

Reporting the outcome in systematically and targeting the treatments cause and financial management. This out generated report represents the final outcome where the scores given to people mark the risk of re admissions in the future or casualties.

Care management and intervention in treatment

Intervening the ongoing treatments and healthcare routines determined by the resultant outcome .Relying on the systems output report the treatment may be altered to prevent the risk predicted.

Patients treatment feedback:

Filing and assessing patient feedbacks and managing of surveys to measure the pros n cons of the analytical system and where the model stands and the future enhancement required.

4. THE PRISM TOOL

PRISM is a probabilistic model inspection tool. Probabilistic model checking is an automatic formal veri¬cation technique for the analysis of systems which exhibit stochastic behaviour.

THE MODEL OF PREDICTIVE RISK STRATIFICATION TOOL (PRISM)

Figure 3 Working of Stratification Model

PRISM model specification

PRISM has direct support for three types of probabilistic models: discrete-time Markov chains (DTMCs), Markov decision processes (MDPs) and continuoustime Markov chains (CTMCs. They are suitable for analysing systems with simple probabilistic behaviour and no concurrency[5].

e.g. synchronous randomised distributed algorithms.

PRISM caculates by performing a series of permutations and combination of nondeterminism and probability, building them to suite modelling multiple probabilistic processes executing in parallel. In some cases where parameters of the system or environmental the behaviour in which it is operating are unknown e.g. component failures and job arrivals[5].

PRISM can also be improved with costs and rewards, real values that manipulate the states and transitions of the model. Thus the reasoning capability off the model is exceeded to atrributes like “completion time”, “energy consumption” or “number of messages lost”.

Models are specified using the PRISM modelling language for the Reactive Modules formalism based on state change. Systems are described modules arranged parallely for processing. Each module’s state is controlled by the assigned probabilistic guarded commands. The language also supports various process algebraic operations with means of global variables and synchronisation. See the PRISM documentation and example repository at [6] for more information.

RESULTANT REPORT

Figure 4 Example to Illustrate the Report of Risk Stratification

4.1 USING PRISM TO STRENGTHEN AND EVALUATE HEALTH INFORMATION SYSTEMS

The PRISM framework identifies strengths and weaknesses in RHIS performance bridging the gaps hence found, leading to the enhancement of health system performance. Routine health information systems (RHIS) try record and present quality information about the health sector organizations. This information is then used as a guide to day-to-day treatments, track routine, rectifying the past results, and hence increasing the accountability.

But the information available in such cases falls short the ideal requirements to produce high quality systems, data quality may be low, intermediate processes of data other may not exist, or managers and staff may have limited knowledge regarding information utility and use of systems, incentives to give attention to the management of information system processes may be few. Looking narrowly at technical issues such as data collection forms we understand the difficulties associated with improving the RHIS systems through PRISM.

PROSPECTS ON CANCER MANAGEMENT

Refined ways to identify and employ multiple, often aggressive, therapies to achieve maximal cancer control its essential to help high-risk patient. Clinician can also give these patients the option to enrol in clinical trials that offer novel therapies. Categorization of patients into established and consistent risk categories is also of key importance in making comparisons between patients in clinical databases.

Sophisticated analytical instruments incorporate risk grouping of similar preoperative clinic pathologic parameters like pre-treatment serum PSA, biopsy score and capacity parameters, and medical tumour stage. Stirring research in the characterization of prostate cancer may one day provide more accurate and individual-specific risk assessment.

First introduced in 1966, Gleason score was introduced to evaluate prostate cancer. In many multivariate cases, the Gleason score proves to be an independent predictor of both pathologic tumours stage and time to biochemical recurrence. Gleason grade may be the most powerful preoperative prognostic factor.

Gleason score as:

• 6 or less as low-risk

• 7 as intermediate-risk

• 8 or above as high-risk

Also, Gleason 7 tumour can be sub classified into either 3+4 or 4+3, depending on which grade is most prevailing in the scores. This category of Gleason score classification and sub classification predicts postoperative outcomes. But the cancer tumour may increase or decrease based on treatment accepted by the recipient it’s a limitation of biopsy Gleason score as a predictor of outcome, its poor correlation with pathologic Gleason score of the surgical specimen Gleason score but still proves to be a good estimator of post operative outcome.

PROSPECT IN CARDIAC ARREST

Preoperative risk scores are a vital tool for risk evaluation, cost-benefit analysis, and preface of new trends. A series of score systems have been developed to predict mortality after performing an adult heart surgery these score systems are based on patient derived data, such as age, gender, and so forth, but there are considerable differences between scores with regard to their design and validity for heart surgery with regard to their predictive values and clinical applicability for our patient population.

Although most of the particular score systems were first and foremost designed to predict mortality, postoperative morbidity has been acknowledged as the major determinant of hospital cost and quality of life after surgery. Therefore, we analyzed the selected risk scores not only with regard to their predictive value for mortality, but for postoperative morbidity as well.

The entire population was then characterized into groups of vague probability of risk of major complications as follows: estimated probability of major cardiac complication <5%, low risk; 5.1% to 25%, medium risk; >25%, high risk. These three sub groups were chosen to provide large enough groups for adequate statistical comparison.

PROSPECT IN DIABETIC HEALTHCARE

Diabetes may be present for up to 7 years before diagnosis early diagnosis, lifestyle modification, and tight glycemic control are necessary to reduce complications; however, these cannot occur if diabetes remains undiagnosed. There is insufficient evidence for or against routine diabetes screening. Reason being the burden and inconvenience caused by fasting visits to meet the diagnostic centres. Diabetes is usually diagnosed by fasting plasma glucose, values which require confirmation on a second visit [7].

Opportunistic programme for high-risk individuals during unscheduled outpatient, urgent care, or hospital visits may improve rates of diagnosis. From the household interview data, we analyzed information on self-reported age, sex, race/ethnicity, education, and income. While providers may choose to use different tools for risk stratification, the principle of deriving a low (<0.5%), moderate (4% to 5%) and high pre-test probability (>10%) could remain similar.[8] Prior reports of diabetes screening in community and clinical venues have yielded mixed results, often limited by low prevalence rates and poor follow-up. Similar to any disease screening, patient adherence with confirmatory testing and subsequent therapy is vital to the successful implementation. Additionally, the cost-effectiveness of opportunistic diabetes screening is unclear and will require further investigation. The proposed algorithm of risk stratification relied on practical reasoning and interpretation of the data; others may suggest thresholds corresponding to different predictive values, and cost effectiveness analysis would further clarify optimal thresholds for clinical practice. Finally, this analysis provides a proposed algorithm, which, if validated, can serve as a guideline for providers, but should not substitute for sound clinical judgment for individual patients.

THE COST ASSOCIATION

Management of the institutes where stratification of health care has been implemented or tested argue the value of disease management programs from a conceptual angle however, most have a difficult time correlating dollars and cents to that value from its practical view point.

As disease management programs have started maturing in size and capacity there exceeds an importance in the task of justifying their expense by demonstrating financial. It is no longer sufficient to defend a program based on an illustrated ROI. Insurers and investors seek in turn the factual relevance, about which members are being identified, the hence taken interventions that can possibly be applied to them with most effectiveness, and which approach leads to genuine performance change and savings. These requirements for data will only amplify in the future, which will lead to insurers and program architects gaining additional concern about economic optimization of disease management efforts. Intensive risk profiling, predictive modelling and stratification will be hence compulsory requirements on the part of all who are involved in program design and execution. [2]

Intervening early reduces costs

Typical high-cost, high-risk disease management program has been administered by a insurer. Members are at high risk because their care is high cost and because they meet definite clinical triggers. Managing these members at the disease stage which may be a non recoverable during this insurer intervenes is largely palliative. In addition, insurer identification methods typically result in a relatively large number of members being referred for management by costly clinical resources.[2]

A more efficient program would identify high-risk members’ earlier prompting intervention with those whose behavior can be changed using risk profiling, prediction and economic modeling for the same.

Summary

As quality control and cost-benefit analysis have gained new relevance with recent developments in the health care system, selection of appropriate score systems for the evaluation of hospital performance has become an important issue to predict and estimate risk scores to predict future admissions and causalities and to ensure health care quality.

Risk stratification is a statistical process by which quality of care can be assessed independently. Evaluation of risk-adjusted patient outcome has become an imperative component of managed care constricting in some markets, and risk-adjusted result rates for hospitals are being reported more frequently in the popular press and on the internet.

The process of risk stratification does not require or assume an extensive arithmetical background. A description of the assumptions for risk stratification provides the quality of various published risk-stratification studies information on evaluation of health care. Numerous practical examples using authentic clinical data help to illustrate risk stratification in health care.

Risk stratification and predictive modelling applications are used in a variety of disease state classification systems derived using claims data. Algorithms based only on pharmacy claims have the recompense of timeliness, hygiene, and accessibility, while still being robust and efficient in the prediction of prospective healthcare outcomes and the costs relative to their incorporated therapeutic and pharmacy counterparts.