Introduction
Clinical order entry and data capture are 2 of the most arduous tasks that physicians face when working with an electronic medical record, or EMR. Of the various technologies available to automate order entry, the use of paper forms that are scanned electronically is the least disruptive to clinicians accustomed to working with traditional charts. Many physicians have resisted using electronic forms on tablet PCs, voice recognition, and other technologies that are used to implement Computerized Physician Order Entry (CPOE) systems.
From a clinical perspective, paper forms that are scanned by nonclinical workers in the background can be minimally disruptive to the clinical practice, especially if there is little or no modification of existing forms.
Moving from a system based on paper charts to an EMR involves dealing with legacy data. For example, even if a physician happens to work in a hospital or clinic that has a full EMR, most patients who were seen by that doctor for even just a few years had some sort of print chart. If only part of a patient's record is online, then many of the advantages of an EMR – including timely access to pertinent data and the ability to search for past diseases and drug allergies – are obviated. One solution is to keep the paper chart around and summarize it as much as possible. The other is to move the pertinent data to the EMR by scanning the paper chart and converting the data to machine-readable and searchable text with Optical Character Recognition (OCR) software.
This article explores clinical data capture options offered by OCR and by Optical Mark Recognition (OMR) software.
OMR Technology
Scanning paper forms to convert marks in checkboxes, text printed in block form, and barcodes into machine readable text is nothing new. The functionality of the technology – flatbed scanners, software, and OMR programs – hasn't evolved significantly in the past decade. However, the cost of the scanners and the PCs needed to run OMR software has decreased significantly in that time.
There are reasons why OMR technology isn't universally employed to create a minimal CPOE. The most significant limitation is that there is no immediate feedback for the ordering physician. An all-electronic system, in contrast, can be designed with an integrated decision support system. However, in many cases these limitations may be acceptable, given the advantages of having the orders automatically entered into the hospital computer system.
I've used Remark's OMR software for about 7 years at Mass General Hospital, Boston. Coupled with a modest HP scanner with sheet feeder, a secretary can use the system to process several dozen forms in the course of 10 minutes. Furthermore, very little hand-holding is required. Once the software is loaded, a stack of papers is fed into the scanner and the software takes over from there. The only minor disruption is not being able to use the PC while forms are being processed.
Optical Character Recognition Technology
One way to move legacy medical record data into a computer system is to scan the image of each page of the record into the computer system. However, the text in the resulting picture of the page isn't recognizable as text to Microsoft Word or most other office applications. OCR software performs the conversion of a bitmapped image to editable text. It's the equivalent of keying in the text by hand, but is less costly and much faster compared with a typical secretary or office worker.
As with OMR software, OCR technology hasn't changed much in the past decade. The programs run faster because the hardware is more advanced, but the basic OCR technology has been stable for years. What has changed is consolidation of the industry. The largest consolidator has been ScanSoft, which acquired TextBridge and its major competitor, OmniPage.
The Process
The process of using a flatbed scanner and OMR or OCR software is straightforward, as shown in the Figure below.
Figure.
A flatbed scanner with a sheet feeder is the rate-limiting step in the conversion of printed documents and forms to machine-readable data.
From the physician's perspective, there is very little to note. A physician fills out a form and drops it off at a collection station (a bin at a nursing station, for example) or inserts it into the chart. A nurse reads the form and acts on the order, and then someone submits the document for scanning. Somewhere in the clinic or hospital, the order entry forms are run through a flatbed scanner, the scanned image is processed by OMR software, and the resulting data are stored in the hospital or clinic database.
Typically, data are stored on a separate server that provides network storage, which allows the data to be incorporated into an EMR and to be accessed by physicians and nurses anywhere on the network. A networked data repository also allows anyone with access privileges to reprint documents that were scanned in and generate new reports from the OMR data.
It's important to note that the combination of flatbed scanner and PC running OMR and/or OCR software simply takes the place of manually typing in the data. As noted earlier, there is no innate clinical data-checking for events such as potential drug-drug interactions. However, it is possible for OMR software to perform some basic checks in the data at the form level. For example, if the physician forgets to check off the route of administration on a drug order, then the OMR software can be programmed to flag the omission. The issue then becomes checking why the order wasn't flagged by the nurse, and determining whether the proper route was used. The process outline in the Figure also assumes that there is someone responsible for the care and feeding of the flatbed scanner and software – which may not be a trivial concern.
An Uncertain Interim
Form-based data entry is probably an interim solution to the clinical data capture problem. The question is, how long is the interim? With full EMR systems in only about 5% of hospitals nationwide, paper records will be with us for the foreseeable future. In time, tablet PCs – or something like them – will replace paper forms, and all orders eventually will be managed by CPOE systems. Until that time, computerizing legacy paper records will be unavoidable if the data are to be searched and manipulated electronically. There will always be that patient transferred from the nursing home with an 8-inch-thick chart.
Another ongoing concern is that OCR is feasible as long as the number of documents to be scanned is modest. Even a higher-end scanner with a sheet feeder and a fast PC can only digitize a few pages per minute. If you have thousands of medical records to digitize, consider a more timely and cost-effective solution, such as outsourcing the translation overseas. Manual rekeying performed in India and Pakistan is cheaper than OCR if you count operator time, hardware, and software costs, even with expedited shipping charges. Of course, there are medical privacy issues to deal with when sending medical records out of your medical institution or practice.
Resources
Commercial-grade flatbed scanners with high-capacity sheet feeders are available for less than $2000. Furthermore, many of these scanners, including models made by HP, come with entry-level OCR programs. The most popular full-featured OCR and OMR software packages run in the $200 to $600 range, respectively. Furthermore, even high-end multi-GHz computers are available for less than $1000.
Of the OCR products on the market, those from ScanSoft are more affordable ($200 or less) and are likely to be found at your local computer shop. Remark OMR is available for around $600. The ABBYY products are higher-end solutions that may be a consideration for large hospitals and clinics. However, whereas ScanSoft products are a one-time cost, ABBYY software OMR and OCR software is leased at a rate based on the number of pages scanned or other use-based licensing.
ABBYY FineReader OCR Software http://www.finereader.com/
Remark Office OMR Software http://www.remark.net/office/
ScanSoft OmniPage Pro OCR Software http://www.scansoft.com/
ScanSoft TextBridge Pro OCR Software http://www.scansoft.com/textbridge/