I had an idea like this about 6-8 months ago. I talked it over with a colleague and they said it couldn’t be done. Well it seems the smart folks at Scorpion Vision Software did what couldn’t be done. Surprise! Denis said they’ve been working on it for about a year. Congratulations are in order as I think this is brilliant.

Translated text from the video description: “This video shows a proof of concept that allows the detection volume of syringes in real time thanks to the vision software. This real-time detection, combined with the reading of bar code products can be an important addition to the quality control of the preparation of intravenous medications in sterile chamber in pharmacies of health facilities.

In this video we demonstrate how the Scorpion vision software software identifies the”bar code” that identifies the type of syringe used, the position of the piston and thesyringe body in 3D space and then estimating the volume contained in the syringe.”

Like many other articles I’ve read recently, this one is from data collected quite a while ago. The information was obtained from 61 patients between  June and October 2010. The article is from the March 2012 issue of the journal. I always marvel at how long it takes study results to get published.

The only problem with the information is that it’s from a 20-month period between January 2005 and August 2006. I hate to break it to you Joint Commission Journal on Quality and Patient Aafety, but that makes the information all but useless. Interesting, but useless.

ISMP does a good job of describing the issue in their March 8, 2012 Medication Safety Alert Newsletter. According to ISMP “proponents of physician dispensing cite improved patient access to medications, patient convenience, greater use of lower-cost generic medications and therapeutic substitutions due to the physician’s enhanced awareness of medication costs, and improved patient adherence with medication regimens. Opponents of physician dispensing cite serious medication safety concerns, particularly the loss of a crucial second check by a pharmacist and use of software to detect prescribing errors, and lack of regulatory oversight, which may lead to lax procedures for medication labeling, record-keeping, storage, and supervision of the dispenser. There’s also a sense of unease regarding a potential conflict of interest when the physician prescribing the medication is also the person dispensing the medication and, perhaps, making a profit from the sale.” That’s a fair assessment of the situation.

My opinion on the subject has changed slightly over the past couple of years. I still believe “remote” dispensing with kiosks can help pharmacist by improving distribution efficiency, but I no longer think it can be done without carefully scrutinizing the motives behind each implementation and how it will best serve the patient.

I think ISMP sums it up best: “ISMP fully supports the removal of any barriers to patients’ access to medications, cost containment that can be achieved by use of lower-cost but effective medications, and steps that improve patient adherence to prescribed medication therapy. However, we cannot support unbridled physician dispensing due to the increased risk of medication errors, particularly with high-alert medications such as chemotherapy. While physician dispensing is permitted in most states, it is often carefully regulated and restricted to samples or conditions of immediate need. Some states require dispensing physicians to be licensed by or registered with the Board of Pharmacy, or they must obtain a special permit. The American Medical Association Code of Ethics notes that “Physicians may dispense drugs within their office practices provided such dispensing primarily benefits the patients.” However, we worry about the unintended consequences of physician dispensing, as described in more detail below.”

I believe in the technology, but I think the circumstances under which we choose to use that technology should be carefully evaluated. Sounds like an opportunity for a good study to me.

Failing to employ available dose error-reduction software (DERS) as intended and to heed important clinical alerts are common contributors to these errors. In particular, the misuse of custom concentration options (i.e., user must fill in the concentration) that do not employ a hard (requires reprogramming) minimum concentration limit is a prime example. This issue contributes largely to preventable errors with smart pumps given the counterintuitive, inverse relationship between concentration and volume. More concentrated drugs require less volume to deliver a specified dose; less concentrated drugs require more volume to deliver a specified dose. When using “fill-in-the-blank” custom concentrations, the concentration must be programmed into the pump so it can calculate the volume needed to deliver the prescribed dose. If the programmed concentration is lower than the actual concentration in the infusion bag or syringe, the pump will deliver an overdose. If the programmed concentration is higher than the actual concentration in the bag or syringe, the pump will deliver an underdose. Without a hard minimum concentration limit, the former scenario has led to life-threatening events, such as those described below.

Here’s a great example from the article that made the light bulb go off in my mind:

A physician prescribed IV HYDROmorphone 20 mg/100 mL (0.2 mg/mL) to infuse at 2.5 mg/hour. In this hospital, the standard concentration for this infusion was 0.1 mg/mL, so the custom concentration of 0.2 mg/mL had to be entered into the smart pump. The nurse selected the custom concentration option then mistakenly entered 2.5 mg/100 mL as the concentration instead of 20 mg/100 mL. Given the erroneously programmed concentration of 0.025 mg/mL, the pump issued a soft (can be overridden) low concentration alert. The nurse overrode the warning, mistakenly believing the warning was inconsequential. Based on the erroneous concentration, the smart pump infused the drug at a rate of 100 mL/hour, while the intended rate was 12.5 mL/hour. The pump delivered the entire bag of HYDROmorphone 20 mg to the patient in 1 hour.

ABSTRACT

Background: Adverse drug reactions (ADRs) are an important source of morbidity and mortality during medical care.
Objective: To examine the trends in mortality related to ADRs reported through the US vital statistics system since January 1999.
Methods: Demographic characteristics of people reported as dying as a result of ADRs from 1999 to 2006 were evaluated. The National Mortality Statistics database was queried for International Classification of Diseases, Tenth Revision, codes Y40-Y59, which are specific for deaths due to adverse effects of drugs in therapeutic use. The data were subgrouped based on demographic factors to identify important trends. Crude rates were calculated based on incidents per 100,000 population. Odds ratios and 95% confidence intervals for subgroups were calculated by logistical regression.
Results: During the 8-year study period 2,313,902,748 person years were evaluated and 2341 ADR-related deaths were identified. Annual rates ranged from 0.08/100,000 to 0.12/100,000, and rates increased significantly over time at a rate of 0.0058 per year. ADR deaths were significantly more likely in persons older than 55 years. The risk was greatest in those aged 75 years or older (OR 6.96, 95% CI 6.30 to 7.69). ADR deaths were higher among men than women. Rates varied by race and ethnicity and were highest among blacks (OR 1.38, 95% CI 1.23 to 1.54). Geographically, rates varied widely between states. Based on urbanization, rates were highest in extremely rural (non-core) areas (OR 2.05, 95% CI 1.76 to 2.38). The most common drug classes associated with death were anticoagulants, opioids, and immunosuppressants.
CONCLUSIONS: ADR death rates have a clear association with age, race, and urbanization subgroups. Older individuals, males, blacks, and individuals residing in extremely rural areas experienced higher ADR death rates; these findings warrant further study to develop prevention strategies.

1. Ann Pharmacother February 2012 vol. 46 no. 2 169-175

Abstract

Satisfactory adherence to aptly prescribed medications is essential for good outcomes of patient care and reliable evaluation of competing modes of drug treatment. The measure of satisfactory adherence is a dosing history that includes timely initiation of dosing plus punctual and persistent execution of the dosing regimen throughout the specified duration of treatment. Standardized terminology for initiation, execution, and persistence of drug dosing is essential for clarity of communication and scientific progress. Electronic methods for compiling drug dosing histories are now the recognized standard for quantifying adherence, the parameters of which support model-based, continuous projections of drug actions and concentrations in plasma that are confirmable by intermittent, direct measurements at single time points. The frequency of inadequate adherence is usually underestimated by pre-electronic methods and thus is clinically unrecognized as a frequent cause of failed treatment or underestimated effectiveness. Intermittent lapses in dosing are potential sources of toxicity through hazardous rebound effects or recurrent first-dose effects.

DrugCite is basically a search engine for the Food and Drug Administrations (FDA) Adverse Event Reporting System (AERS). Acording to the website:

“One of the primary ways that the United States Food and Drug Administration monitors the safety of marketed drugs is the collection and analysis of reported adverse events (an event that was not the intended outcome of the prescribed drug and has a negative impact on health) through the Adverse Events Reporting System (AERS). These reports are submitted by physicians, healthcare consumers, lawyers amongst others, and then the FDA scientific staff will assess these events in the context of other databases to determine if a particular safety concern is associated, and possibly caused by, exposure to a particular drug. Since this is a public database and useful to prescribers and patients alike to know if “has what I’m experiencing been described in patients taking this drug before?” DrugCite has created a more friendly interface to answer that question. Use of this data should in no way be confused with the FDA’s rigorous process of evaluating the potential link of adverse events to drugs as causal but merely represents a portal into what is currently being reported to the Agency.”

It’s a neat concept, but please take the information with a grain of salt. Remember that the FDA will list anything that is reported while an individual is on a medication, whether or not the drug is truly the offending agent or not. For example, I searched for acetaminophen and received the graphs below. Looks pretty scary on the surface, but I consider acetaminophen to be one of the safest medications on the U.S. market when used properly. Something to consider when using this tool.

Several methods have been used to prevent errors associated with high-alert medications. Examples include the use of tallman lettering – which I don’t think actually works – physical separation of the SALAD pairs, the use of brightly-colored warning labels, double checks, etc. All these prevention methods sound promising on the surface, but people are prone to error regardless of the precautions put in place to prevent them. The best we can hope for is to eventually remove humans from the equation.

The Pharmacy Times article recommends the following to help reduce errors associated with high-alert medications:

• Include both brand and generic names, along with indication, when prescribing look- or sound-alike drug names.
• Spell out drug names that have been confused when accepting telephone orders. Require staff to write down the prescription and then perform a read back (and spell back for drugs that are known to cause confusion) of the complete prescription for verification.
• Assign time to provide counseling to patients and/or caregivers, especially for new prescriptions.
• The pharmacist who intercepted the methotrexate error [mentioned in the article] made 2 recommendations that can help catch wrong-drug errors involving highalert medications:
• With all high-alert medications, even if a DUR verification was previously completed, review the prescription’s directions and strength as well as confirm that it is for the right patient before placing it in the bag for pickup.
• Some pharmacy computer systems allow a pharmacist to put a “register hold” on prescriptions so that when the patient picks up the medication they are flagged to speak to a pharmacist. Require an automatic hold to be placed on all high-alert medications so that mandatory counseling occurs, increasing the chance that errors can be discovered. PT

Abstract
Objective To identify the frequency of medication administration errors and their potential risk factors in units using a computerized prescription order entry program and profiled automated dispensing cabinets.

Design Prospective observational study conducted within two clinical units of the Gastroenterology Department in a 1537-bed tertiary teaching hospital in Madrid (Spain).

Measurements Medication errors were measured using the disguised observation technique. Types of medication errors and their potential severity were described. The correlation between potential risk factors and medication errors was studied to identify potential causes.

Results In total, 2314 medication administrations to 73 patients were observed: 509 errors were recorded (22.0%)—68 (13.4%) in preparation and 441 (86.6%) in administration. The most frequent errors were use of wrong administration techniques (especially concerning food intake (13.9%)), wrong reconstitution/dilution (1.7%), omission (1.4%), and wrong infusion speed (1.2%). Errors were classified as no damage (95.7%), no damage but monitoring required (2.3%), and temporary damage (0.4%). Potential clinical severity could not be assessed in 1.6% of cases. The potential risk factors morning shift, evening shift, Anatomical Therapeutic Chemical medication class antacids, prokinetics, antibiotics and immunosuppressants, oral administration, and intravenous administration were associated with a higher risk of administration errors. No association was found with variables related to understaffing or nurse’s experience.

Conclusions Medication administration errors persist in units with automated prescription and dispensing. We identified a need to improve nurses’ working procedures and to implement a Clinical Decision Support tool that generates recommendations about scheduling according to dietary restrictions, preparation of medication before parenteral administration, and adequate infusion rates.

1. J Am Med Inform Assoc. 2012 Jan 1;19(1):72-8. Epub 2011 Sep 2.